Merge branch 'master' into text-embedder-python

WORKSPACE

@@ -177,7 +177,7 @@ http_archive(
         "//third_party:com_google_sentencepiece_no_gflag_no_gtest.diff",
     ],
     patch_args = ["-p1"],
-    repo_mapping = {"@com_google_glog" : "@com_github_glog_glog"},
+    repo_mapping = {"@com_google_glog" : "@com_github_glog_glog_no_gflags"},
 )
 
 http_archive(

docs/build_java_api_docs.py

@@ -45,6 +45,12 @@ def main(_) -> None:
     while (mp_root := mp_root.parent).name != 'mediapipe':
       # Find the nearest `mediapipe` dir.
       pass
+
+    # Externally, parts of the repo are nested inside a mediapipe/ directory
+    # that does not exist internally. Support both.
+    if (mp_root / 'mediapipe').exists():
+      mp_root = mp_root / 'mediapipe'
+
     java_root = mp_root / 'tasks/java'
 
   gen_java.gen_java_docs(

mediapipe/calculators/core/flow_limiter_calculator_test.cc

@@ -79,21 +79,25 @@ std::vector<T> PacketValues(const std::vector<Packet>& packets) {
   return result;
 }
 
-template <typename T>
-std::vector<Packet> MakePackets(std::vector<std::pair<Timestamp, T>> contents) {
-  std::vector<Packet> result;
-  for (auto& entry : contents) {
-    result.push_back(MakePacket<T>(entry.second).At(entry.first));
-  }
-  return result;
-}
-
 std::string SourceString(Timestamp t) {
   return (t.IsSpecialValue())
              ? t.DebugString()
              : absl::StrCat("Timestamp(", t.DebugString(), ")");
 }
 
+template <typename T>
+std::string SourceString(Packet packet) {
+  std::ostringstream oss;
+  if (packet.IsEmpty()) {
+    oss << "Packet()";
+  } else {
+    oss << "MakePacket<" << MediaPipeTypeStringOrDemangled<T>() << ">("
+        << packet.Get<T>() << ")";
+  }
+  oss << ".At(" << SourceString(packet.Timestamp()) << ")";
+  return oss.str();
+}
+
 template <typename PacketContainer, typename PacketContent>
 class PacketsEqMatcher
     : public ::testing::MatcherInterface<const PacketContainer&> {
@@ -123,8 +127,9 @@ class PacketsEqMatcher
     }
     for (auto i1 = c1.begin(), i2 = c2.begin(); i1 != c1.end(); ++i1, ++i2) {
       Packet p1 = *i1, p2 = *i2;
-      if (p1.Timestamp() != p2.Timestamp() ||
+      if (p1.Timestamp() != p2.Timestamp() || p1.IsEmpty() != p2.IsEmpty() ||
-          p1.Get<PacketContent>() != p2.Get<PacketContent>()) {
+          (!p1.IsEmpty() &&
+           p1.Get<PacketContent>() != p2.Get<PacketContent>())) {
         return false;
       }
     }
@@ -133,10 +138,9 @@ class PacketsEqMatcher
   void Print(const PacketContainer& packets, ::std::ostream* os) const {
     for (auto it = packets.begin(); it != packets.end(); ++it) {
       const Packet& packet = *it;
-      *os << (it == packets.begin() ? "{" : "") << "{"
+      *os << (it == packets.begin() ? "{" : "");
-          << SourceString(packet.Timestamp()) << ", "
+      *os << SourceString<PacketContent>(packet);
-          << packet.Get<PacketContent>() << "}"
+      *os << (std::next(it) == packets.end() ? "}" : ", ");
-          << (std::next(it) == packets.end() ? "}" : ", ");
     }
   }
 
@@ -144,7 +148,7 @@ class PacketsEqMatcher
 };
 
 template <typename PacketContainer, typename PacketContent>
-::testing::Matcher<const PacketContainer&> PackestEq(
+::testing::Matcher<const PacketContainer&> PacketsEq(
     const PacketContainer& packets) {
   return MakeMatcher(
       new PacketsEqMatcher<PacketContainer, PacketContent>(packets));
@@ -739,8 +743,8 @@ TEST_F(FlowLimiterCalculatorTest, TwoInputStreams) {
 // The processing time "sleep_time" is reduced from 22ms to 12ms to create
 // the same frame rate as FlowLimiterCalculatorTest::TwoInputStreams.
 TEST_F(FlowLimiterCalculatorTest, ZeroQueue) {
-  auto BoolPackestEq = PackestEq<std::vector<Packet>, bool>;
+  auto BoolPacketsEq = PacketsEq<std::vector<Packet>, bool>;
-  auto IntPackestEq = PackestEq<std::vector<Packet>, int>;
+  auto IntPacketsEq = PacketsEq<std::vector<Packet>, int>;
 
   // Configure the test.
   SetUpInputData();
@@ -835,52 +839,86 @@ TEST_F(FlowLimiterCalculatorTest, ZeroQueue) {
       input_packets_[0],  input_packets_[2],  input_packets_[15],
       input_packets_[17], input_packets_[19],
   };
-  EXPECT_THAT(out_1_packets_, IntPackestEq(expected_output));
+  EXPECT_THAT(out_1_packets_, IntPacketsEq(expected_output));
   // Exactly the timestamps released by FlowLimiterCalculator for in_1_sampled.
   std::vector<Packet> expected_output_2 = {
       input_packets_[0],  input_packets_[2],  input_packets_[4],
       input_packets_[15], input_packets_[17], input_packets_[19],
   };
-  EXPECT_THAT(out_2_packets, IntPackestEq(expected_output_2));
+  EXPECT_THAT(out_2_packets, IntPacketsEq(expected_output_2));
 
   // Validate the ALLOW stream output.
-  std::vector<Packet> expected_allow = MakePackets<bool>(  //
+  std::vector<Packet> expected_allow = {
-      {{Timestamp(0), true},       {Timestamp(10000), false},
+      MakePacket<bool>(true).At(Timestamp(0)),
-       {Timestamp(20000), true},   {Timestamp(30000), false},
+      MakePacket<bool>(false).At(Timestamp(10000)),
-       {Timestamp(40000), true},   {Timestamp(50000), false},
+      MakePacket<bool>(true).At(Timestamp(20000)),
-       {Timestamp(60000), false},  {Timestamp(70000), false},
+      MakePacket<bool>(false).At(Timestamp(30000)),
-       {Timestamp(80000), false},  {Timestamp(90000), false},
+      MakePacket<bool>(true).At(Timestamp(40000)),
-       {Timestamp(100000), false}, {Timestamp(110000), false},
+      MakePacket<bool>(false).At(Timestamp(50000)),
-       {Timestamp(120000), false}, {Timestamp(130000), false},
+      MakePacket<bool>(false).At(Timestamp(60000)),
-       {Timestamp(140000), false}, {Timestamp(150000), true},
+      MakePacket<bool>(false).At(Timestamp(70000)),
-       {Timestamp(160000), false}, {Timestamp(170000), true},
+      MakePacket<bool>(false).At(Timestamp(80000)),
-       {Timestamp(180000), false}, {Timestamp(190000), true},
+      MakePacket<bool>(false).At(Timestamp(90000)),
-       {Timestamp(200000), false}});
+      MakePacket<bool>(false).At(Timestamp(100000)),
-  EXPECT_THAT(allow_packets_, BoolPackestEq(expected_allow));
+      MakePacket<bool>(false).At(Timestamp(110000)),
+      MakePacket<bool>(false).At(Timestamp(120000)),
+      MakePacket<bool>(false).At(Timestamp(130000)),
+      MakePacket<bool>(false).At(Timestamp(140000)),
+      MakePacket<bool>(true).At(Timestamp(150000)),
+      MakePacket<bool>(false).At(Timestamp(160000)),
+      MakePacket<bool>(true).At(Timestamp(170000)),
+      MakePacket<bool>(false).At(Timestamp(180000)),
+      MakePacket<bool>(true).At(Timestamp(190000)),
+      MakePacket<bool>(false).At(Timestamp(200000)),
+  };
+  EXPECT_THAT(allow_packets_, BoolPacketsEq(expected_allow));
+}
+
+std::vector<Packet> StripBoundsUpdates(const std::vector<Packet>& packets,
+                                       Timestamp begin = Timestamp::Min(),
+                                       Timestamp end = Timestamp::Max()) {
+  std::vector<Packet> result;
+  for (const auto& packet : packets) {
+    Timestamp ts = packet.Timestamp();
+    if (packet.IsEmpty() && ts >= begin && ts < end) {
+      continue;
+    }
+    result.push_back(packet);
+  }
+  return result;
 }
 
 // Shows how FlowLimiterCalculator releases auxiliary input packets.
 // In this test, auxiliary input packets arrive at twice the primary rate.
 TEST_F(FlowLimiterCalculatorTest, AuxiliaryInputs) {
-  auto BoolPackestEq = PackestEq<std::vector<Packet>, bool>;
+  auto BoolPacketsEq = PacketsEq<std::vector<Packet>, bool>;
-  auto IntPackestEq = PackestEq<std::vector<Packet>, int>;
+  auto IntPacketsEq = PacketsEq<std::vector<Packet>, int>;
 
   // Configure the test.
   SetUpInputData();
   SetUpSimulationClock();
   CalculatorGraphConfig graph_config =
       ParseTextProtoOrDie<CalculatorGraphConfig>(R"pb(
-        input_stream: 'in_1'
+        input_stream: 'input_1'
-        input_stream: 'in_2'
+        input_stream: 'auxiliary_input_2'
+        input_stream: 'auxiliary_input_3'
         node {
           calculator: 'FlowLimiterCalculator'
-          input_side_packet: 'OPTIONS:limiter_options'
+          options {
-          input_stream: 'in_1'
+            [mediapipe.FlowLimiterCalculatorOptions.ext] {
-          input_stream: 'in_2'
+              max_in_flight: 1
+              max_in_queue: 0
+              in_flight_timeout: 1000000  # 1s
+            }
+          }
+          input_stream: 'input_1'
+          input_stream: 'auxiliary_input_2'
+          input_stream: 'auxiliary_input_3'
           input_stream: 'FINISHED:out_1'
           input_stream_info: { tag_index: 'FINISHED' back_edge: true }
-          output_stream: 'in_1_sampled'
+          output_stream: 'input_1_sampled'
-          output_stream: 'in_2_sampled'
+          output_stream: 'auxiliary_input_2_sampled'
+          output_stream: 'auxiliary_input_3_sampled'
           output_stream: 'ALLOW:allow'
         }
         node {
@@ -888,49 +926,75 @@ TEST_F(FlowLimiterCalculatorTest, AuxiliaryInputs) {
           input_side_packet: 'WARMUP_TIME:warmup_time'
           input_side_packet: 'SLEEP_TIME:sleep_time'
           input_side_packet: 'CLOCK:clock'
-          input_stream: 'PACKET:in_1_sampled'
+          input_stream: 'PACKET:input_1_sampled'
           output_stream: 'PACKET:out_1'
         }
       )pb");
 
-  auto limiter_options = ParseTextProtoOrDie<FlowLimiterCalculatorOptions>(
-      R"pb(
-        max_in_flight: 1 max_in_queue: 0 in_flight_timeout: 1000000  # 1s
-      )pb");
   std::map<std::string, Packet> side_packets = {
-      {"limiter_options",
+      // Fake processing lazy initialization time in microseconds.
-       MakePacket<FlowLimiterCalculatorOptions>(limiter_options)},
       {"warmup_time", MakePacket<int64>(22000)},
+      // Fake processing duration in microseconds.
       {"sleep_time", MakePacket<int64>(22000)},
+      // The SimulationClock to count virtual elapsed time.
       {"clock", MakePacket<mediapipe::Clock*>(clock_)},
   };
 
   // Start the graph.
   MP_ASSERT_OK(graph_.Initialize(graph_config));
-  MP_EXPECT_OK(graph_.ObserveOutputStream("out_1", [this](Packet p) {
+  MP_EXPECT_OK(graph_.ObserveOutputStream(
-    out_1_packets_.push_back(p);
+      "out_1",
-    return absl::OkStatus();
+      [this](Packet p) {
-  }));
+        out_1_packets_.push_back(p);
-  std::vector<Packet> out_2_packets;
+        return absl::OkStatus();
-  MP_EXPECT_OK(graph_.ObserveOutputStream("in_2_sampled", [&](Packet p) {
+      },
-    out_2_packets.push_back(p);
+      true));
-    return absl::OkStatus();
+  std::vector<Packet> out_2_packets, out_3_packets;
-  }));
+  MP_EXPECT_OK(graph_.ObserveOutputStream(
-  MP_EXPECT_OK(graph_.ObserveOutputStream("allow", [this](Packet p) {
+      "auxiliary_input_2_sampled",
-    allow_packets_.push_back(p);
+      [&](Packet p) {
-    return absl::OkStatus();
+        out_2_packets.push_back(p);
-  }));
+        return absl::OkStatus();
+      },
+      true));
+  MP_EXPECT_OK(graph_.ObserveOutputStream(
+      "auxiliary_input_3_sampled",
+      [&](Packet p) {
+        out_3_packets.push_back(p);
+        return absl::OkStatus();
+      },
+      true));
+  MP_EXPECT_OK(graph_.ObserveOutputStream(
+      "allow",
+      [this](Packet p) {
+        allow_packets_.push_back(p);
+        return absl::OkStatus();
+      },
+      true));
   simulation_clock_->ThreadStart();
   MP_ASSERT_OK(graph_.StartRun(side_packets));
 
-  // Add packets 2,4,6,8 to stream in_1 and 1..9 to stream in_2.
+  // Add packets 1..9 to auxiliary_input_3, early.
-  clock_->Sleep(absl::Microseconds(10000));
+  for (int i = 1; i < 10; ++i) {
+    MP_EXPECT_OK(graph_.AddPacketToInputStream(
+        "auxiliary_input_3", MakePacket<int>(i).At(Timestamp(i * 10000))));
+  }
+
+  // The total count of out_2_packets after each input packet.
+  // std::vector<int> sizes_2 = {0, 0, 2, 2, 3, 3, 4, 4, 5, 5};
+  std::vector<int> sizes_2 = {0, 1, 3, 4, 6, 7, 9, 10, 12, 13};
+
+  // Add packets 2,4,6,8 to stream input_1.
+  // Add packets 1..9 to auxiliary_input_2.
   for (int i = 1; i < 10; ++i) {
     if (i % 2 == 0) {
-      MP_EXPECT_OK(graph_.AddPacketToInputStream("in_1", input_packets_[i]));
+      MP_EXPECT_OK(graph_.AddPacketToInputStream(
+          "input_1", MakePacket<int>(i).At(Timestamp(i * 10000))));
     }
-    MP_EXPECT_OK(graph_.AddPacketToInputStream("in_2", input_packets_[i]));
+    MP_EXPECT_OK(graph_.AddPacketToInputStream(
+        "auxiliary_input_2", MakePacket<int>(i).At(Timestamp(i * 10000))));
     clock_->Sleep(absl::Microseconds(10000));
+    EXPECT_EQ(out_2_packets.size(), sizes_2[i]);
   }
 
   // Finish the graph run.
@@ -942,24 +1006,46 @@ TEST_F(FlowLimiterCalculatorTest, AuxiliaryInputs) {
   // Validate the output.
   // Input packets 4 and 8 are dropped due to max_in_flight.
   std::vector<Packet> expected_output = {
-      input_packets_[2],
+      MakePacket<int>(2).At(Timestamp(20000)),
-      input_packets_[6],
+      Packet().At(Timestamp(40000)),
+      MakePacket<int>(6).At(Timestamp(60000)),
+      Packet().At(Timestamp(80000)),
   };
-  EXPECT_THAT(out_1_packets_, IntPackestEq(expected_output));
+  EXPECT_THAT(out_1_packets_, IntPacketsEq(expected_output));
+
   // Packets following input packets 2 and 6, and not input packets 4 and 8.
-  std::vector<Packet> expected_output_2 = {
+  std::vector<Packet> expected_auxiliary_output = {
-      input_packets_[1], input_packets_[2], input_packets_[3],
+      Packet().At(Timestamp(9999)),
-      input_packets_[6], input_packets_[7],
+      MakePacket<int>(1).At(Timestamp(10000)),
+      MakePacket<int>(2).At(Timestamp(20000)),
+      Packet().At(Timestamp(29999)),
+      MakePacket<int>(3).At(Timestamp(30000)),
+      Packet().At(Timestamp(40000)),
+      Packet().At(Timestamp(49999)),
+      Packet().At(Timestamp(50000)),
+      MakePacket<int>(6).At(Timestamp(60000)),
+      Packet().At(Timestamp(69999)),
+      MakePacket<int>(7).At(Timestamp(70000)),
+      Packet().At(Timestamp(80000)),
+      Packet().At(Timestamp(89999)),
   };
-  EXPECT_THAT(out_2_packets, IntPackestEq(expected_output_2));
+  std::vector<Packet> actual_2 =
+      StripBoundsUpdates(out_2_packets, Timestamp(90000));
+  EXPECT_THAT(actual_2, IntPacketsEq(expected_auxiliary_output));
+  std::vector<Packet> expected_3 =
+      StripBoundsUpdates(expected_auxiliary_output, Timestamp(39999));
+  std::vector<Packet> actual_3 =
+      StripBoundsUpdates(out_3_packets, Timestamp(39999));
+  EXPECT_THAT(actual_3, IntPacketsEq(expected_3));
 
   // Validate the ALLOW stream output.
-  std::vector<Packet> expected_allow =
+  std::vector<Packet> expected_allow = {
-      MakePackets<bool>({{Timestamp(20000), 1},
+      MakePacket<bool>(true).At(Timestamp(20000)),
-                         {Timestamp(40000), 0},
+      MakePacket<bool>(false).At(Timestamp(40000)),
-                         {Timestamp(60000), 1},
+      MakePacket<bool>(true).At(Timestamp(60000)),
-                         {Timestamp(80000), 0}});
+      MakePacket<bool>(false).At(Timestamp(80000)),
-  EXPECT_THAT(allow_packets_, BoolPackestEq(expected_allow));
+  };
+  EXPECT_THAT(allow_packets_, BoolPacketsEq(expected_allow));
 }
 
 }  // anonymous namespace

mediapipe/calculators/image/set_alpha_calculator.cc

@@ -46,6 +46,40 @@ constexpr char kOutputFrameTagGpu[] = "IMAGE_GPU";
 constexpr int kNumChannelsRGBA = 4;
 
 enum { ATTRIB_VERTEX, ATTRIB_TEXTURE_POSITION, NUM_ATTRIBUTES };
+
+// Combines an RGB cv::Mat and a single-channel alpha cv::Mat of the same
+// dimensions into an RGBA cv::Mat. Alpha may be read as uint8 or as another
+// numeric type; in the latter case, it is upscaled to values between 0 and 255
+// from an assumed input range of [0, 1). RGB and RGBA Mat's must be uchar.
+template <typename AlphaType>
+absl::Status MergeRGBA8Image(const cv::Mat input_mat, const cv::Mat& alpha_mat,
+                             cv::Mat& output_mat) {
+  RET_CHECK_EQ(input_mat.rows, alpha_mat.rows);
+  RET_CHECK_EQ(input_mat.cols, alpha_mat.cols);
+  RET_CHECK_EQ(input_mat.rows, output_mat.rows);
+  RET_CHECK_EQ(input_mat.cols, output_mat.cols);
+
+  for (int i = 0; i < output_mat.rows; ++i) {
+    const uchar* in_ptr = input_mat.ptr<uchar>(i);
+    const AlphaType* alpha_ptr = alpha_mat.ptr<AlphaType>(i);
+    uchar* out_ptr = output_mat.ptr<uchar>(i);
+    for (int j = 0; j < output_mat.cols; ++j) {
+      const int out_idx = j * kNumChannelsRGBA;
+      const int in_idx = j * input_mat.channels();
+      const int alpha_idx = j * alpha_mat.channels();
+      out_ptr[out_idx + 0] = in_ptr[in_idx + 0];
+      out_ptr[out_idx + 1] = in_ptr[in_idx + 1];
+      out_ptr[out_idx + 2] = in_ptr[in_idx + 2];
+      if constexpr (std::is_same<AlphaType, uchar>::value) {
+        out_ptr[out_idx + 3] = alpha_ptr[alpha_idx + 0];
+      } else {
+        const AlphaType alpha = alpha_ptr[alpha_idx + 0];
+        out_ptr[out_idx + 3] = static_cast<uchar>(round(alpha * 255.0f));
+      }
+    }
+  }
+  return absl::OkStatus();
+}
 }  // namespace
 
 // A calculator for setting the alpha channel of an RGBA image.
@@ -250,28 +284,22 @@ absl::Status SetAlphaCalculator::RenderCpu(CalculatorContext* cc) {
 
   const bool has_alpha_mask = cc->Inputs().HasTag(kInputAlphaTag) &&
                               !cc->Inputs().Tag(kInputAlphaTag).IsEmpty();
-  const bool use_alpa_mask = alpha_value_ < 0 && has_alpha_mask;
+  const bool use_alpha_mask = alpha_value_ < 0 && has_alpha_mask;
 
   // Setup alpha image and Update image in CPU.
-  if (use_alpa_mask) {
+  if (use_alpha_mask) {
     const auto& alpha_mask = cc->Inputs().Tag(kInputAlphaTag).Get<ImageFrame>();
     cv::Mat alpha_mat = mediapipe::formats::MatView(&alpha_mask);
-    RET_CHECK_EQ(input_mat.rows, alpha_mat.rows);
-    RET_CHECK_EQ(input_mat.cols, alpha_mat.cols);
 
-    for (int i = 0; i < output_mat.rows; ++i) {
+    const bool alpha_is_float = alpha_mat.type() == CV_32FC1;
-      const uchar* in_ptr = input_mat.ptr<uchar>(i);
+    RET_CHECK(alpha_is_float || alpha_mat.type() == CV_8UC1);
-      uchar* alpha_ptr = alpha_mat.ptr<uchar>(i);
+
-      uchar* out_ptr = output_mat.ptr<uchar>(i);
+    if (alpha_is_float) {
-      for (int j = 0; j < output_mat.cols; ++j) {
+      MP_RETURN_IF_ERROR(
-        const int out_idx = j * kNumChannelsRGBA;
+          MergeRGBA8Image<float>(input_mat, alpha_mat, output_mat));
-        const int in_idx = j * input_mat.channels();
+    } else {
-        const int alpha_idx = j * alpha_mat.channels();
+      MP_RETURN_IF_ERROR(
-        out_ptr[out_idx + 0] = in_ptr[in_idx + 0];
+          MergeRGBA8Image<uchar>(input_mat, alpha_mat, output_mat));
-        out_ptr[out_idx + 1] = in_ptr[in_idx + 1];
-        out_ptr[out_idx + 2] = in_ptr[in_idx + 2];
-        out_ptr[out_idx + 3] = alpha_ptr[alpha_idx + 0];  // channel 0 of mask
-      }
     }
   } else {
     const uchar alpha_value = std::min(std::max(0.0f, alpha_value_), 255.0f);

mediapipe/calculators/tensor/BUILD

@@ -266,6 +266,7 @@ cc_test(
         "@com_google_absl//absl/status",
         "@com_google_absl//absl/status:statusor",
         "@com_google_absl//absl/strings",
+        "@com_google_sentencepiece//src:sentencepiece_processor",
     ],
 )
 
@@ -321,6 +322,7 @@ cc_test(
         "@com_google_absl//absl/status",
         "@com_google_absl//absl/status:statusor",
         "@com_google_absl//absl/strings",
+        "@com_google_sentencepiece//src:sentencepiece_processor",
     ],
 )
 

mediapipe/calculators/tensorflow/BUILD

@@ -613,6 +613,7 @@ cc_library(
     deps = [
         ":tensorflow_session",
         ":tensorflow_session_from_saved_model_generator_cc_proto",
+        "@com_google_absl//absl/status",
         "//mediapipe/framework:packet_generator",
         "//mediapipe/framework:packet_type",
         "//mediapipe/framework/tool:status_util",

mediapipe/calculators/tensorflow/tensorflow_session_from_saved_model_generator.cc

@@ -14,6 +14,8 @@
 
 #include <algorithm>
 
+#include "absl/status/status.h"
+
 #if !defined(__ANDROID__)
 #include "mediapipe/framework/port/file_helpers.h"
 #endif
@@ -38,6 +40,8 @@ constexpr char kSessionTag[] = "SESSION";
 
 static constexpr char kStringSavedModelPath[] = "STRING_SAVED_MODEL_PATH";
 
+static constexpr char kStringSignatureName[] = "STRING_SIGNATURE_NAME";
+
 // Given the path to a directory containing multiple tensorflow saved models
 // in subdirectories, replaces path with the alphabetically last subdirectory.
 absl::Status GetLatestDirectory(std::string* path) {
@@ -104,6 +108,10 @@ class TensorFlowSessionFromSavedModelGenerator : public PacketGenerator {
     if (input_side_packets->HasTag(kStringSavedModelPath)) {
       input_side_packets->Tag(kStringSavedModelPath).Set<std::string>();
     }
+    // Set Signature_def.
+    if (input_side_packets->HasTag(kStringSignatureName)) {
+      input_side_packets->Tag(kStringSignatureName).Set<std::string>();
+    }
     // A TensorFlow model loaded and ready for use along with tensor
     output_side_packets->Tag(kSessionTag).Set<TensorFlowSession>();
     return absl::OkStatus();
@@ -146,9 +154,19 @@ class TensorFlowSessionFromSavedModelGenerator : public PacketGenerator {
     auto session = absl::make_unique<TensorFlowSession>();
     session->session = std::move(saved_model->session);
 
-    RET_CHECK(!options.signature_name().empty());
+    // Use input side packet to overwrite signature name in options.
+    std::string signature_name =
+        input_side_packets.HasTag(kStringSignatureName)
+            ? input_side_packets.Tag(kStringSignatureName).Get<std::string>()
+            : options.signature_name();
+    RET_CHECK(!signature_name.empty());
     const auto& signature_def_map = saved_model->meta_graph_def.signature_def();
-    const auto& signature_def = signature_def_map.at(options.signature_name());
+    if (signature_def_map.find(signature_name) == signature_def_map.end()) {
+      return absl::NotFoundError(absl::StrFormat(
+          "Signature name '%s' does not exist in the loaded signature def",
+          signature_name));
+    }
+    const auto& signature_def = signature_def_map.at(signature_name);
     for (const auto& input_signature : signature_def.inputs()) {
       session->tag_to_tensor_map[MaybeConvertSignatureToTag(
           input_signature.first, options)] = input_signature.second.name();

mediapipe/calculators/tensorflow/tensorflow_session_from_saved_model_generator_test.cc

@@ -30,11 +30,13 @@
 
 namespace mediapipe {
 
+using ::testing::status::StatusIs;
 namespace {
 
 namespace tf = ::tensorflow;
 
 constexpr char kStringSavedModelPathTag[] = "STRING_SAVED_MODEL_PATH";
+constexpr char kStringSignatureNameTag[] = "STRING_SIGNATURE_NAME";
 constexpr char kSessionTag[] = "SESSION";
 
 std::string GetSavedModelDir() {
@@ -124,6 +126,48 @@ TEST_F(TensorFlowSessionFromSavedModelGeneratorTest,
   ASSERT_NE(session.session, nullptr);
 }
 
+TEST_F(TensorFlowSessionFromSavedModelGeneratorTest,
+       CreateSessionFromSidePacketWithCorrectSignatureName) {
+  generator_options_->clear_saved_model_path();
+  PacketSet input_side_packets(
+      tool::CreateTagMap({"STRING_SAVED_MODEL_PATH:saved_model_dir",
+                          "STRING_SIGNATURE_NAME:signature_name"})
+          .value());
+  input_side_packets.Tag(kStringSavedModelPathTag) =
+      Adopt(new std::string(GetSavedModelDir()));
+  input_side_packets.Tag(kStringSignatureNameTag) =
+      Adopt(new std::string("serving_default"));
+  PacketSet output_side_packets(
+      tool::CreateTagMap({"SESSION:session"}).value());
+  absl::Status run_status = tool::RunGenerateAndValidateTypes(
+      "TensorFlowSessionFromSavedModelGenerator", extendable_options_,
+      input_side_packets, &output_side_packets);
+  MP_EXPECT_OK(run_status) << run_status.message();
+  const TensorFlowSession& session =
+      output_side_packets.Tag(kSessionTag).Get<TensorFlowSession>();
+  // Session must be set.
+  ASSERT_NE(session.session, nullptr);
+}
+
+TEST_F(TensorFlowSessionFromSavedModelGeneratorTest,
+       CreateSessionFromSidePacketWithWrongSignatureName) {
+  generator_options_->clear_saved_model_path();
+  PacketSet input_side_packets(
+      tool::CreateTagMap({"STRING_SAVED_MODEL_PATH:saved_model_dir",
+                          "STRING_SIGNATURE_NAME:signature_name"})
+          .value());
+  input_side_packets.Tag(kStringSavedModelPathTag) =
+      Adopt(new std::string(GetSavedModelDir()));
+  input_side_packets.Tag(kStringSignatureNameTag) =
+      Adopt(new std::string("wrong_signature_name"));
+  PacketSet output_side_packets(
+      tool::CreateTagMap({"SESSION:session"}).value());
+  absl::Status run_status = tool::RunGenerateAndValidateTypes(
+      "TensorFlowSessionFromSavedModelGenerator", extendable_options_,
+      input_side_packets, &output_side_packets);
+  EXPECT_THAT(run_status, StatusIs(absl::StatusCode::kNotFound));
+}
+
 // Integration test. Verifies that TensorFlowInferenceCalculator correctly
 // consumes the Packet emitted by this factory.
 TEST_F(TensorFlowSessionFromSavedModelGeneratorTest,

mediapipe/model_maker/python/text/text_classifier/BUILD

@@ -21,6 +21,19 @@ package(
 
 licenses(["notice"])
 
+py_library(
+    name = "text_classifier_import",
+    srcs = ["__init__.py"],
+    deps = [
+        ":dataset",
+        ":model_options",
+        ":model_spec",
+        ":text_classifier",
+        ":text_classifier_options",
+        "//mediapipe/model_maker/python/core:hyperparameters",
+    ],
+)
+
 py_library(
     name = "model_options",
     srcs = ["model_options.py"],
@@ -114,12 +127,7 @@ py_test(
     ],
     tags = ["requires-net:external"],
     deps = [
-        ":dataset",
+        ":text_classifier_import",
-        ":model_options",
-        ":model_spec",
-        ":text_classifier",
-        ":text_classifier_options",
-        "//mediapipe/model_maker/python/core:hyperparameters",
         "//mediapipe/tasks/python/test:test_utils",
     ],
 )
@@ -128,11 +136,7 @@ py_library(
     name = "text_classifier_demo_lib",
     srcs = ["text_classifier_demo.py"],
     deps = [
-        ":dataset",
+        ":text_classifier_import",
-        ":model_spec",
-        ":text_classifier",
-        ":text_classifier_options",
-        "//mediapipe/model_maker/python/core:hyperparameters",
         "//mediapipe/model_maker/python/core/utils:quantization",
     ],
 )

mediapipe/model_maker/python/text/text_classifier/__init__.py

@@ -11,3 +11,21 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+"""MediaPipe Public Python API for Text Classifier."""
+
+from mediapipe.model_maker.python.core import hyperparameters
+from mediapipe.model_maker.python.text.text_classifier import dataset
+from mediapipe.model_maker.python.text.text_classifier import model_options
+from mediapipe.model_maker.python.text.text_classifier import model_spec
+from mediapipe.model_maker.python.text.text_classifier import text_classifier
+from mediapipe.model_maker.python.text.text_classifier import text_classifier_options
+
+HParams = hyperparameters.BaseHParams
+CSVParams = dataset.CSVParameters
+Dataset = dataset.Dataset
+AverageWordEmbeddingClassifierModelOptions = (
+    model_options.AverageWordEmbeddingClassifierModelOptions)
+BertClassifierModelOptions = model_options.BertClassifierModelOptions
+SupportedModels = model_spec.SupportedModels
+TextClassifier = text_classifier.TextClassifier
+TextClassifierOptions = text_classifier_options.TextClassifierOptions

mediapipe/model_maker/python/text/text_classifier/model_options.py

@@ -18,12 +18,12 @@ from typing import Union
 
 from mediapipe.model_maker.python.text.core import bert_model_options
 
-# BERT text classifier options inherited from BertModelOptions.
+# BERT text classifier model options inherited from BertModelOptions.
-BertClassifierOptions = bert_model_options.BertModelOptions
+BertClassifierModelOptions = bert_model_options.BertModelOptions
 
 
 @dataclasses.dataclass
-class AverageWordEmbeddingClassifierOptions:
+class AverageWordEmbeddingClassifierModelOptions:
   """Configurable model options for an Average Word Embedding classifier.
 
   Attributes:
@@ -41,5 +41,5 @@ class AverageWordEmbeddingClassifierOptions:
   dropout_rate: float = 0.2
 
 
-TextClassifierModelOptions = Union[AverageWordEmbeddingClassifierOptions,
+TextClassifierModelOptions = Union[AverageWordEmbeddingClassifierModelOptions,
-                                   BertClassifierOptions]
+                                   BertClassifierModelOptions]

mediapipe/model_maker/python/text/text_classifier/model_spec.py

@@ -38,8 +38,8 @@ class AverageWordEmbeddingClassifierSpec:
   # `learning_rate` is unused for the average word embedding model
   hparams: hp.BaseHParams = hp.BaseHParams(
       epochs=10, batch_size=32, learning_rate=0)
-  model_options: mo.AverageWordEmbeddingClassifierOptions = (
+  model_options: mo.AverageWordEmbeddingClassifierModelOptions = (
-      mo.AverageWordEmbeddingClassifierOptions())
+      mo.AverageWordEmbeddingClassifierModelOptions())
   name: str = 'AverageWordEmbedding'
 
 

mediapipe/model_maker/python/text/text_classifier/model_spec_test.py

@@ -40,7 +40,7 @@ class ModelSpecTest(tf.test.TestCase):
         })
     self.assertEqual(
         model_spec_obj.model_options,
-        classifier_model_options.BertClassifierOptions(
+        classifier_model_options.BertClassifierModelOptions(
             seq_len=128, do_fine_tuning=True, dropout_rate=0.1))
     self.assertEqual(
         model_spec_obj.hparams,
@@ -57,7 +57,7 @@ class ModelSpecTest(tf.test.TestCase):
     self.assertEqual(model_spec_obj.name, 'AverageWordEmbedding')
     self.assertEqual(
         model_spec_obj.model_options,
-        classifier_model_options.AverageWordEmbeddingClassifierOptions(
+        classifier_model_options.AverageWordEmbeddingClassifierModelOptions(
             seq_len=256,
             wordvec_dim=16,
             do_lower_case=True,
@@ -77,7 +77,7 @@ class ModelSpecTest(tf.test.TestCase):
 
   def test_custom_bert_spec(self):
     custom_bert_classifier_options = (
-        classifier_model_options.BertClassifierOptions(
+        classifier_model_options.BertClassifierModelOptions(
             seq_len=512, do_fine_tuning=False, dropout_rate=0.3))
     model_spec_obj = (
         ms.SupportedModels.MOBILEBERT_CLASSIFIER.value(
@@ -97,7 +97,7 @@ class ModelSpecTest(tf.test.TestCase):
         num_gpus=3,
         tpu='tpu/address')
     custom_average_word_embedding_model_options = (
-        classifier_model_options.AverageWordEmbeddingClassifierOptions(
+        classifier_model_options.AverageWordEmbeddingClassifierModelOptions(
             seq_len=512,
             wordvec_dim=32,
             do_lower_case=False,

mediapipe/model_maker/python/text/text_classifier/text_classifier.py

@@ -51,12 +51,12 @@ def _validate(options: text_classifier_options.TextClassifierOptions):
     return
 
   if (isinstance(options.model_options,
-                 mo.AverageWordEmbeddingClassifierOptions) and
+                 mo.AverageWordEmbeddingClassifierModelOptions) and
       (options.supported_model !=
        ms.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER)):
     raise ValueError("Expected AVERAGE_WORD_EMBEDDING_CLASSIFIER,"
                      f" got {options.supported_model}")
-  if (isinstance(options.model_options, mo.BertClassifierOptions) and
+  if (isinstance(options.model_options, mo.BertClassifierModelOptions) and
       (options.supported_model != ms.SupportedModels.MOBILEBERT_CLASSIFIER)):
     raise ValueError(
         f"Expected MOBILEBERT_CLASSIFIER, got {options.supported_model}")
@@ -194,7 +194,7 @@ class _AverageWordEmbeddingClassifier(TextClassifier):
   _DELIM_REGEX_PATTERN = r"[^\w\']+"
 
   def __init__(self, model_spec: ms.AverageWordEmbeddingClassifierSpec,
-               model_options: mo.AverageWordEmbeddingClassifierOptions,
+               model_options: mo.AverageWordEmbeddingClassifierModelOptions,
                hparams: hp.BaseHParams, label_names: Sequence[str]):
     super().__init__(model_spec, hparams, label_names)
     self._model_options = model_options
@@ -304,8 +304,8 @@ class _BertClassifier(TextClassifier):
   _INITIALIZER_RANGE = 0.02
 
   def __init__(self, model_spec: ms.BertClassifierSpec,
-               model_options: mo.BertClassifierOptions, hparams: hp.BaseHParams,
+               model_options: mo.BertClassifierModelOptions,
-               label_names: Sequence[str]):
+               hparams: hp.BaseHParams, label_names: Sequence[str]):
     super().__init__(model_spec, hparams, label_names)
     self._model_options = model_options
     self._loss_function = tf.keras.losses.SparseCategoricalCrossentropy()

mediapipe/model_maker/python/text/text_classifier/text_classifier_demo.py

@@ -23,12 +23,8 @@ from absl import flags
 from absl import logging
 import tensorflow as tf
 
-from mediapipe.model_maker.python.core import hyperparameters as hp
 from mediapipe.model_maker.python.core.utils import quantization
-from mediapipe.model_maker.python.text.text_classifier import dataset as text_ds
+from mediapipe.model_maker.python.text import text_classifier
-from mediapipe.model_maker.python.text.text_classifier import model_spec as ms
-from mediapipe.model_maker.python.text.text_classifier import text_classifier
-from mediapipe.model_maker.python.text.text_classifier import text_classifier_options
 
 FLAGS = flags.FLAGS
 
@@ -53,31 +49,34 @@ def download_demo_data():
 
 def run(data_dir,
         export_dir=tempfile.mkdtemp(),
-        supported_model=ms.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER):
+        supported_model=(
+            text_classifier.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER)):
   """Runs demo."""
 
   # Gets training data and validation data.
-  csv_params = text_ds.CSVParameters(
+  csv_params = text_classifier.CSVParams(
       text_column='sentence', label_column='label', delimiter='\t')
-  train_data = text_ds.Dataset.from_csv(
+  train_data = text_classifier.Dataset.from_csv(
       filename=os.path.join(os.path.join(data_dir, 'train.tsv')),
       csv_params=csv_params)
-  validation_data = text_ds.Dataset.from_csv(
+  validation_data = text_classifier.Dataset.from_csv(
       filename=os.path.join(os.path.join(data_dir, 'dev.tsv')),
       csv_params=csv_params)
 
   quantization_config = None
-  if supported_model == ms.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER:
+  if (supported_model ==
-    hparams = hp.BaseHParams(
+      text_classifier.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER):
+    hparams = text_classifier.HParams(
         epochs=10, batch_size=32, learning_rate=0, export_dir=export_dir)
   # Warning: This takes extremely long to run on CPU
-  elif supported_model == ms.SupportedModels.MOBILEBERT_CLASSIFIER:
+  elif (
+      supported_model == text_classifier.SupportedModels.MOBILEBERT_CLASSIFIER):
     quantization_config = quantization.QuantizationConfig.for_dynamic()
-    hparams = hp.BaseHParams(
+    hparams = text_classifier.HParams(
         epochs=3, batch_size=48, learning_rate=3e-5, export_dir=export_dir)
 
   # Fine-tunes the model.
-  options = text_classifier_options.TextClassifierOptions(
+  options = text_classifier.TextClassifierOptions(
       supported_model=supported_model, hparams=hparams)
   model = text_classifier.TextClassifier.create(train_data, validation_data,
                                                 options)
@@ -96,9 +95,10 @@ def main(_):
   export_dir = os.path.expanduser(FLAGS.export_dir)
 
   if FLAGS.supported_model == 'average_word_embedding':
-    supported_model = ms.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER
+    supported_model = (
+        text_classifier.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER)
   elif FLAGS.supported_model == 'bert':
-    supported_model = ms.SupportedModels.MOBILEBERT_CLASSIFIER
+    supported_model = text_classifier.SupportedModels.MOBILEBERT_CLASSIFIER
 
   run(data_dir, export_dir, supported_model)
 

mediapipe/model_maker/python/text/text_classifier/text_classifier_test.py

@@ -18,12 +18,7 @@ import os
 
 import tensorflow as tf
 
-from mediapipe.model_maker.python.core import hyperparameters as hp
+from mediapipe.model_maker.python.text import text_classifier
-from mediapipe.model_maker.python.text.text_classifier import dataset
-from mediapipe.model_maker.python.text.text_classifier import model_options as mo
-from mediapipe.model_maker.python.text.text_classifier import model_spec as ms
-from mediapipe.model_maker.python.text.text_classifier import text_classifier
-from mediapipe.model_maker.python.text.text_classifier import text_classifier_options
 from mediapipe.tasks.python.test import test_utils
 
 
@@ -43,18 +38,23 @@ class TextClassifierTest(tf.test.TestCase):
       writer.writeheader()
       for label, text in labels_and_text:
         writer.writerow({'text': text, 'label': label})
-    csv_params = dataset.CSVParameters(text_column='text', label_column='label')
+    csv_params = text_classifier.CSVParams(
-    all_data = dataset.Dataset.from_csv(
+        text_column='text', label_column='label')
+    all_data = text_classifier.Dataset.from_csv(
         filename=csv_file, csv_params=csv_params)
     return all_data.split(0.5)
 
   def test_create_and_train_average_word_embedding_model(self):
     train_data, validation_data = self._get_data()
-    options = text_classifier_options.TextClassifierOptions(
+    options = (
-        supported_model=ms.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER,
+        text_classifier.TextClassifierOptions(
-        hparams=hp.BaseHParams(epochs=1, batch_size=1, learning_rate=0))
+            supported_model=(text_classifier.SupportedModels
-    average_word_embedding_classifier = text_classifier.TextClassifier.create(
+                             .AVERAGE_WORD_EMBEDDING_CLASSIFIER),
-        train_data, validation_data, options)
+            hparams=text_classifier.HParams(
+                epochs=1, batch_size=1, learning_rate=0)))
+    average_word_embedding_classifier = (
+        text_classifier.TextClassifier.create(train_data, validation_data,
+                                              options))
 
     _, accuracy = average_word_embedding_classifier.evaluate(validation_data)
     self.assertGreaterEqual(accuracy, 0.0)
@@ -77,10 +77,11 @@ class TextClassifierTest(tf.test.TestCase):
 
   def test_create_and_train_bert(self):
     train_data, validation_data = self._get_data()
-    options = text_classifier_options.TextClassifierOptions(
+    options = text_classifier.TextClassifierOptions(
-        supported_model=ms.SupportedModels.MOBILEBERT_CLASSIFIER,
+        supported_model=text_classifier.SupportedModels.MOBILEBERT_CLASSIFIER,
-        model_options=mo.BertClassifierOptions(do_fine_tuning=False, seq_len=2),
+        model_options=text_classifier.BertClassifierModelOptions(
-        hparams=hp.BaseHParams(
+            do_fine_tuning=False, seq_len=2),
+        hparams=text_classifier.HParams(
             epochs=1,
             batch_size=1,
             learning_rate=3e-5,
@@ -94,12 +95,13 @@ class TextClassifierTest(tf.test.TestCase):
 
   def test_label_mismatch(self):
     options = (
-        text_classifier_options.TextClassifierOptions(
+        text_classifier.TextClassifierOptions(
-            supported_model=ms.SupportedModels.MOBILEBERT_CLASSIFIER))
+            supported_model=(
+                text_classifier.SupportedModels.MOBILEBERT_CLASSIFIER)))
     train_tf_dataset = tf.data.Dataset.from_tensor_slices([[0]])
-    train_data = dataset.Dataset(train_tf_dataset, 1, ['foo'])
+    train_data = text_classifier.Dataset(train_tf_dataset, 1, ['foo'])
     validation_tf_dataset = tf.data.Dataset.from_tensor_slices([[0]])
-    validation_data = dataset.Dataset(validation_tf_dataset, 1, ['bar'])
+    validation_data = text_classifier.Dataset(validation_tf_dataset, 1, ['bar'])
     with self.assertRaisesRegex(
         ValueError,
         'Training data label names .* not equal to validation data label names'
@@ -111,9 +113,11 @@ class TextClassifierTest(tf.test.TestCase):
     train_data, validation_data = self._get_data()
 
     avg_options = (
-        text_classifier_options.TextClassifierOptions(
+        text_classifier.TextClassifierOptions(
-            supported_model=ms.SupportedModels.MOBILEBERT_CLASSIFIER,
+            supported_model=(
-            model_options=mo.AverageWordEmbeddingClassifierOptions()))
+                text_classifier.SupportedModels.MOBILEBERT_CLASSIFIER),
+            model_options=(
+                text_classifier.AverageWordEmbeddingClassifierModelOptions())))
     with self.assertRaisesRegex(
         ValueError, 'Expected AVERAGE_WORD_EMBEDDING_CLASSIFIER, got'
         ' SupportedModels.MOBILEBERT_CLASSIFIER'):
@@ -121,10 +125,10 @@ class TextClassifierTest(tf.test.TestCase):
                                             avg_options)
 
     bert_options = (
-        text_classifier_options.TextClassifierOptions(
+        text_classifier.TextClassifierOptions(
-            supported_model=(
+            supported_model=(text_classifier.SupportedModels
-                ms.SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER),
+                             .AVERAGE_WORD_EMBEDDING_CLASSIFIER),
-            model_options=mo.BertClassifierOptions()))
+            model_options=text_classifier.BertClassifierModelOptions()))
     with self.assertRaisesRegex(
         ValueError, 'Expected MOBILEBERT_CLASSIFIER, got'
         ' SupportedModels.AVERAGE_WORD_EMBEDDING_CLASSIFIER'):

mediapipe/python/BUILD

@@ -89,6 +89,7 @@ cc_library(
     deps = [
         "//mediapipe/tasks/cc/vision/gesture_recognizer:gesture_recognizer_graph",
         "//mediapipe/tasks/cc/vision/image_classifier:image_classifier_graph",
+        "//mediapipe/tasks/cc/vision/image_embedder:image_embedder_graph",
         "//mediapipe/tasks/cc/vision/image_segmenter:image_segmenter_graph",
         "//mediapipe/tasks/cc/vision/object_detector:object_detector_graph",
     ] + select({

mediapipe/tasks/cc/text/text_classifier/BUILD

@@ -85,6 +85,7 @@ cc_test(
         "@com_google_absl//absl/status:statusor",
         "@com_google_absl//absl/strings",
         "@com_google_absl//absl/strings:cord",
+        "@com_google_sentencepiece//src:sentencepiece_processor",
         "@org_tensorflow//tensorflow/lite/core/shims:cc_shims_test_util",
     ],
 )

mediapipe/tasks/cc/text/text_embedder/BUILD

@@ -82,6 +82,7 @@ cc_test(
         "@com_google_absl//absl/flags:flag",
         "@com_google_absl//absl/status",
         "@com_google_absl//absl/status:statusor",
+        "@com_google_sentencepiece//src:sentencepiece_processor",
         "@org_tensorflow//tensorflow/lite/core/shims:cc_shims_test_util",
     ],
 )

mediapipe/tasks/cc/text/tokenizers/BUILD

@@ -83,6 +83,7 @@ cc_test(
         ":sentencepiece_tokenizer",
         "//mediapipe/framework/port:gtest_main",
         "//mediapipe/tasks/cc/core:utils",
+        "@com_google_sentencepiece//src:sentencepiece_processor",
     ],
 )
 
@@ -132,6 +133,7 @@ cc_test(
         "@com_google_absl//absl/status:statusor",
         "@com_google_absl//absl/strings",
         "@com_google_absl//absl/strings:cord",
+        "@com_google_sentencepiece//src:sentencepiece_processor",
     ],
 )
 

mediapipe/tasks/python/components/containers/embedding_result.py

@@ -14,7 +14,7 @@
 """Embeddings data class."""
 
 import dataclasses
-from typing import Any, Optional, List
+from typing import Optional, List
 
 import numpy as np
 from mediapipe.tasks.cc.components.containers.proto import embeddings_pb2
@@ -36,29 +36,11 @@ class FloatEmbedding:
 
   values: np.ndarray
 
-  @doc_controls.do_not_generate_docs
-  def to_pb2(self) -> _FloatEmbeddingProto:
-    """Generates a FloatEmbedding protobuf object."""
-    return _FloatEmbeddingProto(values=self.values)
-
   @classmethod
   @doc_controls.do_not_generate_docs
-  def create_from_pb2(
+  def create_from_pb2(cls, pb2_obj: _FloatEmbeddingProto) -> 'FloatEmbedding':
-      cls, pb2_obj: _FloatEmbeddingProto) -> 'FloatEmbedding':
     """Creates a `FloatEmbedding` object from the given protobuf object."""
-    return FloatEmbedding(values=np.array(pb2_obj.value_float, dtype=float))
+    return FloatEmbedding(values=np.array(pb2_obj.values, dtype=float))
-
-  def __eq__(self, other: Any) -> bool:
-    """Checks if this object is equal to the given object.
-    Args:
-      other: The object to be compared with.
-    Returns:
-      True if the objects are equal.
-    """
-    if not isinstance(other, FloatEmbedding):
-      return False
-
-    return self.to_pb2().__eq__(other.to_pb2())
 
 
 @dataclasses.dataclass
@@ -71,30 +53,13 @@ class QuantizedEmbedding:
 
   values: np.ndarray
 
-  @doc_controls.do_not_generate_docs
-  def to_pb2(self) -> _QuantizedEmbeddingProto:
-    """Generates a QuantizedEmbedding protobuf object."""
-    return _QuantizedEmbeddingProto(values=self.values)
-
   @classmethod
   @doc_controls.do_not_generate_docs
   def create_from_pb2(
       cls, pb2_obj: _QuantizedEmbeddingProto) -> 'QuantizedEmbedding':
     """Creates a `QuantizedEmbedding` object from the given protobuf object."""
     return QuantizedEmbedding(
-        values=np.array(bytearray(pb2_obj.value_string), dtype=np.uint8))
+        values=np.array(bytearray(pb2_obj.values), dtype=np.uint8))
-
-  def __eq__(self, other: Any) -> bool:
-    """Checks if this object is equal to the given object.
-    Args:
-      other: The object to be compared with.
-    Returns:
-      True if the objects are equal.
-    """
-    if not isinstance(other, QuantizedEmbedding):
-      return False
-
-    return self.to_pb2().__eq__(other.to_pb2())
 
 
 @dataclasses.dataclass
@@ -113,58 +78,31 @@ class Embedding:
   head_index: Optional[int] = None
   head_name: Optional[str] = None
 
-  @doc_controls.do_not_generate_docs
-  def to_pb2(self) -> _EmbeddingProto:
-    """Generates a Embedding protobuf object."""
-
-    if self.embedding.dtype == float:
-      return _EmbeddingProto(float_embedding=self.embedding,
-                             head_index=self.head_index,
-                             head_name=self.head_name)
-
-    elif self.embedding.dtype == np.uint8:
-      return _EmbeddingProto(quantized_embedding=bytes(self.embedding),
-                             head_index=self.head_index,
-                             head_name=self.head_name)
-
-    else:
-      raise ValueError("Invalid dtype. Only float and np.uint8 are supported.")
-
   @classmethod
   @doc_controls.do_not_generate_docs
-  def create_from_pb2(
+  def create_from_pb2(cls, pb2_obj: _EmbeddingProto) -> 'Embedding':
-      cls, pb2_obj: _EmbeddingProto) -> 'Embedding':
     """Creates a `Embedding` object from the given protobuf object."""
 
     quantized_embedding = np.array(
         bytearray(pb2_obj.quantized_embedding.values))
     float_embedding = np.array(pb2_obj.float_embedding.values, dtype=float)
 
-    if len(quantized_embedding) == 0:
+    if not quantized_embedding:
-      return Embedding(embedding=float_embedding,
+      return Embedding(
-                       head_index=pb2_obj.head_index,
+          embedding=float_embedding,
-                       head_name=pb2_obj.head_name)
+          head_index=pb2_obj.head_index,
+          head_name=pb2_obj.head_name)
     else:
-      return Embedding(embedding=quantized_embedding,
+      return Embedding(
-                       head_index=pb2_obj.head_index,
+          embedding=quantized_embedding,
-                       head_name=pb2_obj.head_name)
+          head_index=pb2_obj.head_index,
-
+          head_name=pb2_obj.head_name)
-  def __eq__(self, other: Any) -> bool:
-    """Checks if this object is equal to the given object.
-    Args:
-      other: The object to be compared with.
-    Returns:
-      True if the objects are equal.
-    """
-    if not isinstance(other, Embedding):
-      return False
-
-    return self.to_pb2().__eq__(other.to_pb2())
 
 
 @dataclasses.dataclass
 class EmbeddingResult:
   """Embedding results for a given embedder model.
+
   Attributes:
     embeddings: A list of `Embedding` objects.
     timestamp_ms: The optional timestamp (in milliseconds) of the start of the
@@ -178,33 +116,10 @@ class EmbeddingResult:
   embeddings: List[Embedding]
   timestamp_ms: Optional[int] = None
 
-  @doc_controls.do_not_generate_docs
-  def to_pb2(self) -> _EmbeddingResultProto:
-    """Generates a EmbeddingResult protobuf object."""
-    return _EmbeddingResultProto(
-        embeddings=[
-            embedding.to_pb2() for embedding in self.embeddings
-        ])
-
   @classmethod
   @doc_controls.do_not_generate_docs
-  def create_from_pb2(
+  def create_from_pb2(cls, pb2_obj: _EmbeddingResultProto) -> 'EmbeddingResult':
-      cls, pb2_obj: _EmbeddingResultProto) -> 'EmbeddingResult':
     """Creates a `EmbeddingResult` object from the given protobuf object."""
-    return EmbeddingResult(
+    return EmbeddingResult(embeddings=[
-        embeddings=[
+        Embedding.create_from_pb2(embedding) for embedding in pb2_obj.embeddings
-            Embedding.create_from_pb2(embedding)
+    ])
-            for embedding in pb2_obj.embeddings
-        ])
-
-  def __eq__(self, other: Any) -> bool:
-    """Checks if this object is equal to the given object.
-    Args:
-      other: The object to be compared with.
-    Returns:
-      True if the objects are equal.
-    """
-    if not isinstance(other, EmbeddingResult):
-      return False
-
-    return self.to_pb2().__eq__(other.to_pb2())

mediapipe/tasks/python/components/processors/embedder_options.py

@@ -44,16 +44,14 @@ class EmbedderOptions:
   def to_pb2(self) -> _EmbedderOptionsProto:
     """Generates a EmbedderOptions protobuf object."""
     return _EmbedderOptionsProto(
-        l2_normalize=self.l2_normalize,
+        l2_normalize=self.l2_normalize, quantize=self.quantize)
-        quantize=self.quantize)
 
   @classmethod
   @doc_controls.do_not_generate_docs
   def create_from_pb2(cls, pb2_obj: _EmbedderOptionsProto) -> 'EmbedderOptions':
     """Creates a `EmbedderOptions` object from the given protobuf object."""
     return EmbedderOptions(
-        l2_normalize=pb2_obj.l2_normalize,
+        l2_normalize=pb2_obj.l2_normalize, quantize=pb2_obj.quantize)
-        quantize=pb2_obj.quantize)
 
   def __eq__(self, other: Any) -> bool:
     """Checks if this object is equal to the given object.

mediapipe/tasks/python/components/utils/BUILD

@@ -14,6 +14,8 @@
 
 # Placeholder for internal Python strict library compatibility macro.
 
+# Placeholder for internal Python strict library and test compatibility macro.
+
 package(default_visibility = ["//mediapipe/tasks:internal"])
 
 licenses(["notice"])

mediapipe/tasks/python/components/utils/cosine_similarity.py

@@ -23,6 +23,8 @@ _EmbedderOptions = embedder_options.EmbedderOptions
 
 
 def _compute_cosine_similarity(u, v):
+  """Computes cosine similarity between two embeddings."""
+
   if len(u.embedding) <= 0:
     raise ValueError("Cannot compute cosing similarity on empty embeddings.")
 
@@ -38,6 +40,7 @@ def _compute_cosine_similarity(u, v):
 
 def cosine_similarity(u: _Embedding, v: _Embedding) -> float:
   """Utility function to compute cosine similarity between two embedding.
+
   May return an InvalidArgumentError if e.g. the feature vectors are of
   different types (quantized vs. float), have different sizes, or have an
   L2-norm of 0.
@@ -45,6 +48,9 @@ def cosine_similarity(u: _Embedding, v: _Embedding) -> float:
   Args:
     u: An embedding.
     v: An embedding.
+
+  Returns:
+    Cosine similarity value.
   """
   if len(u.embedding) != len(v.embedding):
     raise ValueError(f"Cannot compute cosine similarity between embeddings "

mediapipe/tasks/python/test/text/text_classifier_test.py

@@ -125,7 +125,7 @@ class ModelFileType(enum.Enum):
   FILE_NAME = 2
 
 
-class ImageClassifierTest(parameterized.TestCase):
+class TextClassifierTest(parameterized.TestCase):
 
   def setUp(self):
     super().setUp()

mediapipe/tasks/python/test/vision/BUILD

@@ -58,6 +58,26 @@ py_test(
     ],
 )
 
+py_test(
+    name = "image_embedder_test",
+    srcs = ["image_embedder_test.py"],
+    data = [
+        "//mediapipe/tasks/testdata/vision:test_images",
+        "//mediapipe/tasks/testdata/vision:test_models",
+    ],
+    deps = [
+        "//mediapipe/python:_framework_bindings",
+        "//mediapipe/tasks/python/components/containers:embedding_result",
+        "//mediapipe/tasks/python/components/containers:rect",
+        "//mediapipe/tasks/python/components/processors:embedder_options",
+        "//mediapipe/tasks/python/core:base_options",
+        "//mediapipe/tasks/python/test:test_utils",
+        "//mediapipe/tasks/python/vision:image_embedder",
+        "//mediapipe/tasks/python/vision/core:image_processing_options",
+        "//mediapipe/tasks/python/vision/core:vision_task_running_mode",
+    ],
+)
+
 py_test(
     name = "image_segmenter_test",
     srcs = ["image_segmenter_test.py"],

mediapipe/tasks/python/test/vision/image_embedder_test.py

@@ -0,0 +1,402 @@
+# Copyright 2022 The MediaPipe Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Tests for image embedder."""
+
+import enum
+import os
+from unittest import mock
+
+from absl.testing import absltest
+from absl.testing import parameterized
+import numpy as np
+
+from mediapipe.python._framework_bindings import image as image_module
+from mediapipe.tasks.python.components.containers import embedding_result as embedding_result_module
+from mediapipe.tasks.python.components.containers import rect
+from mediapipe.tasks.python.components.processors import embedder_options as embedder_options_module
+from mediapipe.tasks.python.core import base_options as base_options_module
+from mediapipe.tasks.python.test import test_utils
+from mediapipe.tasks.python.vision import image_embedder
+from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module
+from mediapipe.tasks.python.vision.core import vision_task_running_mode as running_mode_module
+
+ImageEmbedderResult = embedding_result_module.EmbeddingResult
+_Rect = rect.Rect
+_BaseOptions = base_options_module.BaseOptions
+_EmbedderOptions = embedder_options_module.EmbedderOptions
+_FloatEmbedding = embedding_result_module.FloatEmbedding
+_QuantizedEmbedding = embedding_result_module.QuantizedEmbedding
+_Embedding = embedding_result_module.Embedding
+_Image = image_module.Image
+_ImageEmbedder = image_embedder.ImageEmbedder
+_ImageEmbedderOptions = image_embedder.ImageEmbedderOptions
+_RUNNING_MODE = running_mode_module.VisionTaskRunningMode
+_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
+
+_MODEL_FILE = 'mobilenet_v3_small_100_224_embedder.tflite'
+_BURGER_IMAGE_FILE = 'burger.jpg'
+_BURGER_CROPPED_IMAGE_FILE = 'burger_crop.jpg'
+_TEST_DATA_DIR = 'mediapipe/tasks/testdata/vision'
+# Tolerance for embedding vector coordinate values.
+_EPSILON = 1e-4
+# Tolerance for cosine similarity evaluation.
+_SIMILARITY_TOLERANCE = 1e-6
+
+
+class ModelFileType(enum.Enum):
+  FILE_CONTENT = 1
+  FILE_NAME = 2
+
+
+class ImageEmbedderTest(parameterized.TestCase):
+
+  def setUp(self):
+    super().setUp()
+    self.test_image = _Image.create_from_file(
+        test_utils.get_test_data_path(
+            os.path.join(_TEST_DATA_DIR, _BURGER_IMAGE_FILE)))
+    self.test_cropped_image = _Image.create_from_file(
+        test_utils.get_test_data_path(
+            os.path.join(_TEST_DATA_DIR, _BURGER_CROPPED_IMAGE_FILE)))
+    self.model_path = test_utils.get_test_data_path(
+        os.path.join(_TEST_DATA_DIR, _MODEL_FILE))
+
+  def test_create_from_file_succeeds_with_valid_model_path(self):
+    # Creates with default option and valid model file successfully.
+    with _ImageEmbedder.create_from_model_path(self.model_path) as embedder:
+      self.assertIsInstance(embedder, _ImageEmbedder)
+
+  def test_create_from_options_succeeds_with_valid_model_path(self):
+    # Creates with options containing model file successfully.
+    base_options = _BaseOptions(model_asset_path=self.model_path)
+    options = _ImageEmbedderOptions(base_options=base_options)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      self.assertIsInstance(embedder, _ImageEmbedder)
+
+  def test_create_from_options_fails_with_invalid_model_path(self):
+    with self.assertRaisesRegex(
+        RuntimeError, 'Unable to open file at /path/to/invalid/model.tflite'):
+      base_options = _BaseOptions(
+          model_asset_path='/path/to/invalid/model.tflite')
+      options = _ImageEmbedderOptions(base_options=base_options)
+      _ImageEmbedder.create_from_options(options)
+
+  def test_create_from_options_succeeds_with_valid_model_content(self):
+    # Creates with options containing model content successfully.
+    with open(self.model_path, 'rb') as f:
+      base_options = _BaseOptions(model_asset_buffer=f.read())
+      options = _ImageEmbedderOptions(base_options=base_options)
+      embedder = _ImageEmbedder.create_from_options(options)
+      self.assertIsInstance(embedder, _ImageEmbedder)
+
+  def _check_embedding_value(self, result, expected_first_value):
+    # Check embedding first value.
+    self.assertAlmostEqual(
+        result.embeddings[0].embedding[0], expected_first_value, delta=_EPSILON)
+
+  def _check_embedding_size(self, result, quantize, expected_embedding_size):
+    # Check embedding size.
+    self.assertLen(result.embeddings, 1)
+    embedding_result = result.embeddings[0]
+    self.assertLen(embedding_result.embedding, expected_embedding_size)
+    if quantize:
+      self.assertEqual(embedding_result.embedding.dtype, np.uint8)
+    else:
+      self.assertEqual(embedding_result.embedding.dtype, float)
+
+  def _check_cosine_similarity(self, result0, result1, expected_similarity):
+    # Checks cosine similarity.
+    similarity = _ImageEmbedder.cosine_similarity(result0.embeddings[0],
+                                                  result1.embeddings[0])
+    self.assertAlmostEqual(
+        similarity, expected_similarity, delta=_SIMILARITY_TOLERANCE)
+
+  @parameterized.parameters(
+      (False, False, False, ModelFileType.FILE_NAME, 0.925519, 1024,
+       (-0.2101883, -0.193027)),
+      (True, False, False, ModelFileType.FILE_NAME, 0.925519, 1024,
+       (-0.0142344, -0.0131606)),
+      # (False, True, False, ModelFileType.FILE_NAME,
+      #  0.926791, 1024, (229, 231)),
+      (False, False, True, ModelFileType.FILE_CONTENT, 0.999931, 1024,
+       (-0.195062, -0.193027)))
+  def test_embed(self, l2_normalize, quantize, with_roi, model_file_type,
+                 expected_similarity, expected_size, expected_first_values):
+    # Creates embedder.
+    if model_file_type is ModelFileType.FILE_NAME:
+      base_options = _BaseOptions(model_asset_path=self.model_path)
+    elif model_file_type is ModelFileType.FILE_CONTENT:
+      with open(self.model_path, 'rb') as f:
+        model_content = f.read()
+      base_options = _BaseOptions(model_asset_buffer=model_content)
+    else:
+      # Should never happen
+      raise ValueError('model_file_type is invalid.')
+
+    embedder_options = _EmbedderOptions(
+        l2_normalize=l2_normalize, quantize=quantize)
+    options = _ImageEmbedderOptions(
+        base_options=base_options, embedder_options=embedder_options)
+    embedder = _ImageEmbedder.create_from_options(options)
+
+    image_processing_options = None
+    if with_roi:
+      # Region-of-interest in "burger.jpg" corresponding to "burger_crop.jpg".
+      roi = _Rect(left=0, top=0, right=0.833333, bottom=1)
+      image_processing_options = _ImageProcessingOptions(roi)
+
+    # Extracts both embeddings.
+    image_result = embedder.embed(self.test_image, image_processing_options)
+    crop_result = embedder.embed(self.test_cropped_image)
+
+    # Checks embeddings and cosine similarity.
+    expected_result0_value, expected_result1_value = expected_first_values
+    self._check_embedding_size(image_result, quantize, expected_size)
+    self._check_embedding_size(crop_result, quantize, expected_size)
+    self._check_embedding_value(image_result, expected_result0_value)
+    self._check_embedding_value(crop_result, expected_result1_value)
+    self._check_cosine_similarity(image_result, crop_result,
+                                  expected_similarity)
+    # Closes the embedder explicitly when the embedder is not used in
+    # a context.
+    embedder.close()
+
+  @parameterized.parameters(
+      (False, False, ModelFileType.FILE_NAME, 0.925519),
+      (False, False, ModelFileType.FILE_CONTENT, 0.925519))
+  def test_embed_in_context(self, l2_normalize, quantize, model_file_type,
+                            expected_similarity):
+    # Creates embedder.
+    if model_file_type is ModelFileType.FILE_NAME:
+      base_options = _BaseOptions(model_asset_path=self.model_path)
+    elif model_file_type is ModelFileType.FILE_CONTENT:
+      with open(self.model_path, 'rb') as f:
+        model_content = f.read()
+      base_options = _BaseOptions(model_asset_buffer=model_content)
+    else:
+      # Should never happen
+      raise ValueError('model_file_type is invalid.')
+
+    embedder_options = _EmbedderOptions(
+        l2_normalize=l2_normalize, quantize=quantize)
+    options = _ImageEmbedderOptions(
+        base_options=base_options, embedder_options=embedder_options)
+
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      # Extracts both embeddings.
+      image_result = embedder.embed(self.test_image)
+      crop_result = embedder.embed(self.test_cropped_image)
+
+      # Checks cosine similarity.
+      self._check_cosine_similarity(image_result, crop_result,
+                                    expected_similarity)
+
+  def test_missing_result_callback(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM)
+    with self.assertRaisesRegex(ValueError,
+                                r'result callback must be provided'):
+      with _ImageEmbedder.create_from_options(options) as unused_embedder:
+        pass
+
+  @parameterized.parameters((_RUNNING_MODE.IMAGE), (_RUNNING_MODE.VIDEO))
+  def test_illegal_result_callback(self, running_mode):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=running_mode,
+        result_callback=mock.MagicMock())
+    with self.assertRaisesRegex(ValueError,
+                                r'result callback should not be provided'):
+      with _ImageEmbedder.create_from_options(options) as unused_embedder:
+        pass
+
+  def test_calling_embed_for_video_in_image_mode(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      with self.assertRaisesRegex(ValueError,
+                                  r'not initialized with the video mode'):
+        embedder.embed_for_video(self.test_image, 0)
+
+  def test_calling_embed_async_in_image_mode(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      with self.assertRaisesRegex(ValueError,
+                                  r'not initialized with the live stream mode'):
+        embedder.embed_async(self.test_image, 0)
+
+  def test_calling_embed_in_video_mode(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      with self.assertRaisesRegex(ValueError,
+                                  r'not initialized with the image mode'):
+        embedder.embed(self.test_image)
+
+  def test_calling_embed_async_in_video_mode(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      with self.assertRaisesRegex(ValueError,
+                                  r'not initialized with the live stream mode'):
+        embedder.embed_async(self.test_image, 0)
+
+  def test_embed_for_video_with_out_of_order_timestamp(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      unused_result = embedder.embed_for_video(self.test_image, 1)
+      with self.assertRaisesRegex(
+          ValueError, r'Input timestamp must be monotonically increasing'):
+        embedder.embed_for_video(self.test_image, 0)
+
+  def test_embed_for_video(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO)
+    with _ImageEmbedder.create_from_options(options) as embedder0, \
+         _ImageEmbedder.create_from_options(options) as embedder1:
+      for timestamp in range(0, 300, 30):
+        # Extracts both embeddings.
+        image_result = embedder0.embed_for_video(self.test_image, timestamp)
+        crop_result = embedder1.embed_for_video(self.test_cropped_image,
+                                                timestamp)
+        # Checks cosine similarity.
+        self._check_cosine_similarity(
+            image_result, crop_result, expected_similarity=0.925519)
+
+  def test_embed_for_video_succeeds_with_region_of_interest(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.VIDEO)
+    with _ImageEmbedder.create_from_options(options) as embedder0, \
+         _ImageEmbedder.create_from_options(options) as embedder1:
+      # Region-of-interest in "burger.jpg" corresponding to "burger_crop.jpg".
+      roi = _Rect(left=0, top=0, right=0.833333, bottom=1)
+      image_processing_options = _ImageProcessingOptions(roi)
+
+      for timestamp in range(0, 300, 30):
+        # Extracts both embeddings.
+        image_result = embedder0.embed_for_video(self.test_image, timestamp,
+                                                 image_processing_options)
+        crop_result = embedder1.embed_for_video(self.test_cropped_image,
+                                                timestamp)
+
+        # Checks cosine similarity.
+        self._check_cosine_similarity(
+            image_result, crop_result, expected_similarity=0.999931)
+
+  def test_calling_embed_in_live_stream_mode(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock())
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      with self.assertRaisesRegex(ValueError,
+                                  r'not initialized with the image mode'):
+        embedder.embed(self.test_image)
+
+  def test_calling_embed_for_video_in_live_stream_mode(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock())
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      with self.assertRaisesRegex(ValueError,
+                                  r'not initialized with the video mode'):
+        embedder.embed_for_video(self.test_image, 0)
+
+  def test_embed_async_calls_with_illegal_timestamp(self):
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=mock.MagicMock())
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      embedder.embed_async(self.test_image, 100)
+      with self.assertRaisesRegex(
+          ValueError, r'Input timestamp must be monotonically increasing'):
+        embedder.embed_async(self.test_image, 0)
+
+  def test_embed_async_calls(self):
+    # Get the embedding result for the cropped image.
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      crop_result = embedder.embed(self.test_cropped_image)
+
+    observed_timestamp_ms = -1
+
+    def check_result(result: ImageEmbedderResult, output_image: _Image,
+                     timestamp_ms: int):
+      # Checks cosine similarity.
+      self._check_cosine_similarity(
+          result, crop_result, expected_similarity=0.925519)
+      self.assertTrue(
+          np.array_equal(output_image.numpy_view(),
+                         self.test_image.numpy_view()))
+      self.assertLess(observed_timestamp_ms, timestamp_ms)
+      self.observed_timestamp_ms = timestamp_ms
+
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=check_result)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      for timestamp in range(0, 300, 30):
+        embedder.embed_async(self.test_image, timestamp)
+
+  def test_embed_async_succeeds_with_region_of_interest(self):
+    # Get the embedding result for the cropped image.
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.IMAGE)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      crop_result = embedder.embed(self.test_cropped_image)
+
+    # Region-of-interest in "burger.jpg" corresponding to "burger_crop.jpg".
+    roi = _Rect(left=0, top=0, right=0.833333, bottom=1)
+    image_processing_options = _ImageProcessingOptions(roi)
+    observed_timestamp_ms = -1
+
+    def check_result(result: ImageEmbedderResult, output_image: _Image,
+                     timestamp_ms: int):
+      # Checks cosine similarity.
+      self._check_cosine_similarity(
+          result, crop_result, expected_similarity=0.999931)
+      self.assertTrue(
+          np.array_equal(output_image.numpy_view(),
+                         self.test_image.numpy_view()))
+      self.assertLess(observed_timestamp_ms, timestamp_ms)
+      self.observed_timestamp_ms = timestamp_ms
+
+    options = _ImageEmbedderOptions(
+        base_options=_BaseOptions(model_asset_path=self.model_path),
+        running_mode=_RUNNING_MODE.LIVE_STREAM,
+        result_callback=check_result)
+    with _ImageEmbedder.create_from_options(options) as embedder:
+      for timestamp in range(0, 300, 30):
+        embedder.embed_async(self.test_image, timestamp,
+                             image_processing_options)
+
+
+if __name__ == '__main__':
+  absltest.main()

mediapipe/tasks/python/vision/BUILD

@@ -79,6 +79,29 @@ py_library(
     ],
 )
 
+py_library(
+    name = "image_embedder",
+    srcs = [
+        "image_embedder.py",
+    ],
+    deps = [
+        "//mediapipe/python:_framework_bindings",
+        "//mediapipe/python:packet_creator",
+        "//mediapipe/python:packet_getter",
+        "//mediapipe/tasks/cc/components/containers/proto:embeddings_py_pb2",
+        "//mediapipe/tasks/cc/vision/image_embedder/proto:image_embedder_graph_options_py_pb2",
+        "//mediapipe/tasks/python/components/containers:embedding_result",
+        "//mediapipe/tasks/python/components/processors:embedder_options",
+        "//mediapipe/tasks/python/components/utils:cosine_similarity",
+        "//mediapipe/tasks/python/core:base_options",
+        "//mediapipe/tasks/python/core:optional_dependencies",
+        "//mediapipe/tasks/python/core:task_info",
+        "//mediapipe/tasks/python/vision/core:base_vision_task_api",
+        "//mediapipe/tasks/python/vision/core:image_processing_options",
+        "//mediapipe/tasks/python/vision/core:vision_task_running_mode",
+    ],
+)
+
 py_library(
     name = "gesture_recognizer",
     srcs = [

mediapipe/tasks/python/vision/__init__.py

@@ -18,6 +18,7 @@ import mediapipe.tasks.python.vision.core
 import mediapipe.tasks.python.vision.gesture_recognizer
 import mediapipe.tasks.python.vision.hand_landmarker
 import mediapipe.tasks.python.vision.image_classifier
+import mediapipe.tasks.python.vision.image_embedder
 import mediapipe.tasks.python.vision.image_segmenter
 import mediapipe.tasks.python.vision.object_detector
 
@@ -29,6 +30,10 @@ HandLandmarkerOptions = hand_landmarker.HandLandmarkerOptions
 HandLandmarkerResult = hand_landmarker.HandLandmarkerResult
 ImageClassifier = image_classifier.ImageClassifier
 ImageClassifierOptions = image_classifier.ImageClassifierOptions
+ImageClassifierResult = image_classifier.ImageClassifierResult
+ImageEmbedder = image_embedder.ImageEmbedder
+ImageEmbedderOptions = image_embedder.ImageEmbedderOptions
+ImageEmbedderResult = image_embedder.ImageEmbedderResult
 ImageSegmenter = image_segmenter.ImageSegmenter
 ImageSegmenterOptions = image_segmenter.ImageSegmenterOptions
 ObjectDetector = object_detector.ObjectDetector
@@ -40,6 +45,7 @@ del core
 del gesture_recognizer
 del hand_landmarker
 del image_classifier
+del image_embedder
 del image_segmenter
 del object_detector
 del mediapipe

mediapipe/tasks/python/vision/image_embedder.py

@@ -0,0 +1,309 @@
+# Copyright 2022 The MediaPipe Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""MediaPipe image embedder task."""
+
+import dataclasses
+from typing import Callable, Mapping, Optional
+
+from mediapipe.python import packet_creator
+from mediapipe.python import packet_getter
+from mediapipe.python._framework_bindings import image as image_module
+from mediapipe.python._framework_bindings import packet as packet_module
+from mediapipe.tasks.cc.components.containers.proto import embeddings_pb2
+from mediapipe.tasks.cc.vision.image_embedder.proto import image_embedder_graph_options_pb2
+from mediapipe.tasks.python.components.containers import embedding_result as embedding_result_module
+from mediapipe.tasks.python.components.processors import embedder_options
+from mediapipe.tasks.python.components.utils import cosine_similarity
+from mediapipe.tasks.python.core import base_options as base_options_module
+from mediapipe.tasks.python.core import task_info as task_info_module
+from mediapipe.tasks.python.core.optional_dependencies import doc_controls
+from mediapipe.tasks.python.vision.core import base_vision_task_api
+from mediapipe.tasks.python.vision.core import image_processing_options as image_processing_options_module
+from mediapipe.tasks.python.vision.core import vision_task_running_mode as running_mode_module
+
+ImageEmbedderResult = embedding_result_module.EmbeddingResult
+_BaseOptions = base_options_module.BaseOptions
+_ImageEmbedderGraphOptionsProto = image_embedder_graph_options_pb2.ImageEmbedderGraphOptions
+_EmbedderOptions = embedder_options.EmbedderOptions
+_RunningMode = running_mode_module.VisionTaskRunningMode
+_TaskInfo = task_info_module.TaskInfo
+_ImageProcessingOptions = image_processing_options_module.ImageProcessingOptions
+
+_EMBEDDINGS_OUT_STREAM_NAME = 'embeddings_out'
+_EMBEDDINGS_TAG = 'EMBEDDINGS'
+_IMAGE_IN_STREAM_NAME = 'image_in'
+_IMAGE_OUT_STREAM_NAME = 'image_out'
+_IMAGE_TAG = 'IMAGE'
+_NORM_RECT_STREAM_NAME = 'norm_rect_in'
+_NORM_RECT_TAG = 'NORM_RECT'
+_TASK_GRAPH_NAME = 'mediapipe.tasks.vision.image_embedder.ImageEmbedderGraph'
+_MICRO_SECONDS_PER_MILLISECOND = 1000
+
+
+@dataclasses.dataclass
+class ImageEmbedderOptions:
+  """Options for the image embedder task.
+
+  Attributes:
+    base_options: Base options for the image embedder task.
+    running_mode: The running mode of the task. Default to the image mode. Image
+      embedder task has three running modes: 1) The image mode for embedding
+      image on single image inputs. 2) The video mode for embedding image on the
+      decoded frames of a video. 3) The live stream mode for embedding image on
+      a live stream of input data, such as from camera.
+    embedder_options: Options for the image embedder task.
+    result_callback: The user-defined result callback for processing live stream
+      data. The result callback should only be specified when the running mode
+      is set to the live stream mode.
+  """
+  base_options: _BaseOptions
+  running_mode: _RunningMode = _RunningMode.IMAGE
+  embedder_options: _EmbedderOptions = _EmbedderOptions()
+  result_callback: Optional[Callable[
+      [ImageEmbedderResult, image_module.Image, int], None]] = None
+
+  @doc_controls.do_not_generate_docs
+  def to_pb2(self) -> _ImageEmbedderGraphOptionsProto:
+    """Generates an ImageEmbedderOptions protobuf object."""
+    base_options_proto = self.base_options.to_pb2()
+    base_options_proto.use_stream_mode = False if self.running_mode == _RunningMode.IMAGE else True
+    embedder_options_proto = self.embedder_options.to_pb2()
+
+    return _ImageEmbedderGraphOptionsProto(
+        base_options=base_options_proto,
+        embedder_options=embedder_options_proto)
+
+
+class ImageEmbedder(base_vision_task_api.BaseVisionTaskApi):
+  """Class that performs embedding extraction on images."""
+
+  @classmethod
+  def create_from_model_path(cls, model_path: str) -> 'ImageEmbedder':
+    """Creates an `ImageEmbedder` object from a TensorFlow Lite model and the default `ImageEmbedderOptions`.
+
+    Note that the created `ImageEmbedder` instance is in image mode, for
+    embedding image on single image inputs.
+
+    Args:
+      model_path: Path to the model.
+
+    Returns:
+      `ImageEmbedder` object that's created from the model file and the default
+      `ImageEmbedderOptions`.
+
+    Raises:
+      ValueError: If failed to create `ImageEmbedder` object from the provided
+        file such as invalid file path.
+      RuntimeError: If other types of error occurred.
+    """
+    base_options = _BaseOptions(model_asset_path=model_path)
+    options = ImageEmbedderOptions(
+        base_options=base_options, running_mode=_RunningMode.IMAGE)
+    return cls.create_from_options(options)
+
+  @classmethod
+  def create_from_options(cls,
+                          options: ImageEmbedderOptions) -> 'ImageEmbedder':
+    """Creates the `ImageEmbedder` object from image embedder options.
+
+    Args:
+      options: Options for the image embedder task.
+
+    Returns:
+      `ImageEmbedder` object that's created from `options`.
+
+    Raises:
+      ValueError: If failed to create `ImageEmbedder` object from
+        `ImageEmbedderOptions` such as missing the model.
+      RuntimeError: If other types of error occurred.
+    """
+
+    def packets_callback(output_packets: Mapping[str, packet_module.Packet]):
+      if output_packets[_IMAGE_OUT_STREAM_NAME].is_empty():
+        return
+
+      embedding_result_proto = embeddings_pb2.EmbeddingResult()
+      embedding_result_proto.CopyFrom(
+          packet_getter.get_proto(output_packets[_EMBEDDINGS_OUT_STREAM_NAME]))
+
+      image = packet_getter.get_image(output_packets[_IMAGE_OUT_STREAM_NAME])
+      timestamp = output_packets[_IMAGE_OUT_STREAM_NAME].timestamp
+      options.result_callback(
+          ImageEmbedderResult.create_from_pb2(embedding_result_proto), image,
+          timestamp.value // _MICRO_SECONDS_PER_MILLISECOND)
+
+    task_info = _TaskInfo(
+        task_graph=_TASK_GRAPH_NAME,
+        input_streams=[
+            ':'.join([_IMAGE_TAG, _IMAGE_IN_STREAM_NAME]),
+            ':'.join([_NORM_RECT_TAG, _NORM_RECT_STREAM_NAME]),
+        ],
+        output_streams=[
+            ':'.join([_EMBEDDINGS_TAG, _EMBEDDINGS_OUT_STREAM_NAME]),
+            ':'.join([_IMAGE_TAG, _IMAGE_OUT_STREAM_NAME])
+        ],
+        task_options=options)
+    return cls(
+        task_info.generate_graph_config(
+            enable_flow_limiting=options.running_mode ==
+            _RunningMode.LIVE_STREAM), options.running_mode,
+        packets_callback if options.result_callback else None)
+
+  def embed(
+      self,
+      image: image_module.Image,
+      image_processing_options: Optional[_ImageProcessingOptions] = None
+  ) -> ImageEmbedderResult:
+    """Performs image embedding extraction on the provided MediaPipe Image.
+
+     Extraction is performed on the region of interest specified by the `roi`
+     argument if provided, or on the entire image otherwise.
+
+    Args:
+      image: MediaPipe Image.
+      image_processing_options: Options for image processing.
+
+    Returns:
+      An embedding result object that contains a list of embeddings.
+
+    Raises:
+      ValueError: If any of the input arguments is invalid.
+      RuntimeError: If image embedder failed to run.
+    """
+    normalized_rect = self.convert_to_normalized_rect(image_processing_options)
+    output_packets = self._process_image_data({
+        _IMAGE_IN_STREAM_NAME:
+            packet_creator.create_image(image),
+        _NORM_RECT_STREAM_NAME:
+            packet_creator.create_proto(normalized_rect.to_pb2())
+    })
+
+    embedding_result_proto = embeddings_pb2.EmbeddingResult()
+    embedding_result_proto.CopyFrom(
+        packet_getter.get_proto(output_packets[_EMBEDDINGS_OUT_STREAM_NAME]))
+
+    return ImageEmbedderResult.create_from_pb2(embedding_result_proto)
+
+  def embed_for_video(
+      self,
+      image: image_module.Image,
+      timestamp_ms: int,
+      image_processing_options: Optional[_ImageProcessingOptions] = None
+  ) -> ImageEmbedderResult:
+    """Performs image embedding extraction on the provided video frames.
+
+    Extraction is performed on the region of interested specified by the `roi`
+    argument if provided, or on the entire image otherwise.
+
+    Only use this method when the ImageEmbedder is created with the video
+    running mode. It's required to provide the video frame's timestamp (in
+    milliseconds) along with the video frame. The input timestamps should be
+    monotonically increasing for adjacent calls of this method.
+
+    Args:
+      image: MediaPipe Image.
+      timestamp_ms: The timestamp of the input video frame in milliseconds.
+      image_processing_options: Options for image processing.
+
+    Returns:
+      An embedding result object that contains a list of embeddings.
+
+    Raises:
+      ValueError: If any of the input arguments is invalid.
+      RuntimeError: If image embedder failed to run.
+    """
+    normalized_rect = self.convert_to_normalized_rect(image_processing_options)
+    output_packets = self._process_video_data({
+        _IMAGE_IN_STREAM_NAME:
+            packet_creator.create_image(image).at(
+                timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND),
+        _NORM_RECT_STREAM_NAME:
+            packet_creator.create_proto(normalized_rect.to_pb2()).at(
+                timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND)
+    })
+    embedding_result_proto = embeddings_pb2.EmbeddingResult()
+    embedding_result_proto.CopyFrom(
+        packet_getter.get_proto(output_packets[_EMBEDDINGS_OUT_STREAM_NAME]))
+
+    return ImageEmbedderResult.create_from_pb2(embedding_result_proto)
+
+  def embed_async(
+      self,
+      image: image_module.Image,
+      timestamp_ms: int,
+      image_processing_options: Optional[_ImageProcessingOptions] = None
+  ) -> None:
+    """Sends live image data to embedder.
+
+    The results will be available via the "result_callback" provided in the
+    ImageEmbedderOptions. Embedding extraction is performed on the region of
+    interested specified by the `roi` argument if provided, or on the entire
+    image otherwise.
+
+    Only use this method when the ImageEmbedder is created with the live
+    stream running mode. The input timestamps should be monotonically increasing
+    for adjacent calls of this method. This method will return immediately after
+    the input image is accepted. The results will be available via the
+    `result_callback` provided in the `ImageEmbedderOptions`. The
+    `embed_async` method is designed to process live stream data such as
+    camera input. To lower the overall latency, image embedder may drop the
+    input images if needed. In other words, it's not guaranteed to have output
+    per input image.
+
+    The `result_callback` provides:
+      - An embedding result object that contains a list of embeddings.
+      - The input image that the image embedder runs on.
+      - The input timestamp in milliseconds.
+
+    Args:
+      image: MediaPipe Image.
+      timestamp_ms: The timestamp of the input image in milliseconds.
+      image_processing_options: Options for image processing.
+
+    Raises:
+      ValueError: If the current input timestamp is smaller than what the image
+        embedder has already processed.
+    """
+    normalized_rect = self.convert_to_normalized_rect(image_processing_options)
+    self._send_live_stream_data({
+        _IMAGE_IN_STREAM_NAME:
+            packet_creator.create_image(image).at(
+                timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND),
+        _NORM_RECT_STREAM_NAME:
+            packet_creator.create_proto(normalized_rect.to_pb2()).at(
+                timestamp_ms * _MICRO_SECONDS_PER_MILLISECOND)
+    })
+
+  @classmethod
+  def cosine_similarity(cls, u: embedding_result_module.Embedding,
+                        v: embedding_result_module.Embedding) -> float:
+    """Utility function to compute cosine similarity between two embedding entries.
+
+    May return an InvalidArgumentError if e.g. the feature vectors are of
+    different types (quantized vs. float), have different sizes, or have an
+    L2-norm of 0.
+
+    Args:
+      u: An embedding entry.
+      v: An embedding entry.
+
+    Returns:
+      The cosine similarity for the two embeddings.
+
+    Raises:
+      ValueError: May return an error if e.g. the feature vectors are of
+        different types (quantized vs. float), have different sizes, or have
+        an L2-norm of 0.
+    """
+    return cosine_similarity.cosine_similarity(u, v)

mediapipe/tasks/web/BUILD

@@ -31,6 +31,7 @@ rollup_bundle(
     name = "audio_bundle",
     config_file = "rollup.config.mjs",
     entry_point = "audio.ts",
+    format = "cjs",
     output_dir = False,
     deps = [
         ":audio_lib",
@@ -67,6 +68,7 @@ rollup_bundle(
     name = "text_bundle",
     config_file = "rollup.config.mjs",
     entry_point = "text.ts",
+    format = "cjs",
     output_dir = False,
     deps = [
         ":text_lib",
@@ -103,6 +105,7 @@ rollup_bundle(
     name = "vision_bundle",
     config_file = "rollup.config.mjs",
     entry_point = "vision.ts",
+    format = "cjs",
     output_dir = False,
     deps = [
         ":vision_lib",

mediapipe/tasks/web/package.json

@@ -2,16 +2,15 @@
   "name": "@mediapipe/tasks-__NAME__",
   "version": "__VERSION__",
   "description": "__DESCRIPTION__",
-  "main": "__NAME__bundle.js",
+  "main": "__NAME___bundle.js",
-  "module":  "__NAME__bundle.js",
+  "module":  "__NAME___bundle.js",
   "exports": {
-    ".": "./__NAME__bundle.js",
+    ".": "./__NAME___bundle.js",
-    "./loader": "./wasm/__NAME__wasm_internal.js",
+    "./loader": "./wasm/__NAME___wasm_internal.js",
-    "./wasm": "./wasm/__NAME__wasm_internal.wasm"
+    "./wasm": "./wasm/__NAME___wasm_internal.wasm"
   },
   "author": "mediapipe@google.com",
   "license": "Apache-2.0",
-  "type": "module",
   "dependencies": {
     "google-protobuf": "^3.21.2"
   },

mediapipe/util/android/asset_manager_util.h

@@ -32,17 +32,23 @@ namespace mediapipe {
 
 // Thin wrapper over AAssetManager provided by JNI. This class is meant to be
 // used as a singleton.
-// Usage: Call InitializeFromActivity from a JNI function that has access to the
+//
-// Java activity in the Android application. This initializes the asset manager
+// Usage: Call one of Initialize* functions from a JNI function that has access
-// and now files bundled in the assets folder can be read using ReadFile().
+// to a context/activity/etc. This initializes the asset manager and now files
+// bundled in the assets folder can be read using ReadFile().
+//
+// NOTE: initialization should happen strictly once and guaranteed to complete
+// before any possible use, otherwise it cannot be used safely across multiple
+// threads.
 class AssetManager {
  public:
   AssetManager(const AssetManager&) = delete;
   AssetManager& operator=(const AssetManager&) = delete;
 
-  // Returns the asset manager if it has been set by a call to
+  // Returns the asset manager if it has been set by one of Initialize*
-  // InitializeFromActivity, otherwise returns nullptr.
+  // functions, otherwise returns nullptr.
   AAssetManager* GetAssetManager();
+
   // Returns true if AAssetManager was successfully initialized.
   bool InitializeFromAssetManager(JNIEnv* env, jobject local_asset_manager,
                                   const std::string& cache_dir_path);
@@ -55,7 +61,7 @@ class AssetManager {
                               const std::string& cache_dir_path);
 
   // Returns true if AAssetManager was successfully initialized.
-  ABSL_DEPRECATED("Use InitializeFromActivity instead.")
+  ABSL_DEPRECATED("Use one of alternate Initialize* functions instead.")
   bool InitializeFromAssetManager(JNIEnv* env, jobject local_asset_manager);
 
   // Returns true if AAssetManager was successfully initialized.
@@ -79,12 +85,14 @@ class AssetManager {
                               std::string* output);
 
   // Returns the path to the Android cache directory. Will be empty if
-  // InitializeFromActivity has not been called.
+  // AssetManager hasn't been initialized.
   const std::string& GetCacheDirPath();
 
   // Caches the contents of the given asset as a file, and returns a path to
   // that file. This can be used to pass an asset to APIs that require a path
   // to a filesystem file.
+  // NOTE: this is _not_ thread-safe, e.g. if two threads are requesting the
+  // same file
   absl::StatusOr<std::string> CachedFileFromAsset(
       const std::string& asset_path);