Project import generated by Copybara.

GitOrigin-RevId: ba1d851bc868c2f8037a6fa96ee90e4b8ab9bd40

WORKSPACE

@@ -191,7 +191,7 @@ maven_install(
         "androidx.recyclerview:recyclerview:aar:1.1.0-beta02",
         "com.google.android.material:material:aar:1.0.0-rc01",
     ],
-   repositories = [
+    repositories = [
         "https://dl.google.com/dl/android/maven2",
         "https://repo1.maven.org/maven2",
     ],
@@ -210,10 +210,10 @@ maven_jar(
 )
 
 maven_jar(
-     name = "androidx_concurrent_futures",
-     artifact = "androidx.concurrent:concurrent-futures:1.0.0-alpha03",
-     sha1 = "b528df95c7e2fefa2210c0c742bf3e491c1818ae",
-     server = "google_server",
+    name = "androidx_concurrent_futures",
+    artifact = "androidx.concurrent:concurrent-futures:1.0.0-alpha03",
+    sha1 = "b528df95c7e2fefa2210c0c742bf3e491c1818ae",
+    server = "google_server",
 )
 
 maven_jar(

mediapipe/calculators/core/previous_loopback_calculator.cc

@@ -74,10 +74,6 @@ class PreviousLoopbackCalculator : public CalculatorBase {
   }
 
   ::mediapipe::Status Process(CalculatorContext* cc) final {
-    Packet& main_packet = cc->Inputs().Get(main_id_).Value();
-    if (!main_packet.IsEmpty()) {
-      main_ts_.push_back(main_packet.Timestamp());
-    }
     Packet& loopback_packet = cc->Inputs().Get(loop_id_).Value();
     if (!loopback_packet.IsEmpty()) {
       loopback_packets_.push_back(loopback_packet);
@@ -87,6 +83,23 @@ class PreviousLoopbackCalculator : public CalculatorBase {
       }
     }
 
+    Packet& main_packet = cc->Inputs().Get(main_id_).Value();
+    if (!main_packet.IsEmpty()) {
+      main_ts_.push_back(main_packet.Timestamp());
+
+      // In case of an empty "LOOP" input, truncate timestamp is set to the
+      // lowest possible timestamp for a successive non-empty "LOOP" input. This
+      // truncates main_ts_ as soon as possible, and produces the highest legal
+      // output timestamp bound.
+      if (loopback_packet.IsEmpty() &&
+          loopback_packet.Timestamp() != Timestamp::Unstarted()) {
+        while (!main_ts_.empty() &&
+               main_ts_.front() <= loopback_packet.Timestamp() + 1) {
+          main_ts_.pop_front();
+        }
+      }
+    }
+
     while (!main_ts_.empty() && !loopback_packets_.empty()) {
       Timestamp main_timestamp = main_ts_.front();
       main_ts_.pop_front();

mediapipe/calculators/core/previous_loopback_calculator_test.cc

@@ -198,5 +198,64 @@ TEST(PreviousLoopbackCalculator, ClosesCorrectly) {
   MP_EXPECT_OK(graph_.WaitUntilDone());
 }
 
+// Demonstrates that downstream calculators won't be blocked by
+// always-empty-LOOP-stream.
+TEST(PreviousLoopbackCalculator, EmptyLoopForever) {
+  std::vector<Packet> outputs;
+  CalculatorGraphConfig graph_config_ =
+      ParseTextProtoOrDie<CalculatorGraphConfig>(R"(
+        input_stream: 'in'
+        node {
+          calculator: 'PreviousLoopbackCalculator'
+          input_stream: 'MAIN:in'
+          input_stream: 'LOOP:previous'
+          input_stream_info: { tag_index: 'LOOP' back_edge: true }
+          output_stream: 'PREV_LOOP:previous'
+        }
+        # This calculator synchronizes its inputs as normal, so it is used
+        # to check that both "in" and "previous" are ready.
+        node {
+          calculator: 'PassThroughCalculator'
+          input_stream: 'in'
+          input_stream: 'previous'
+          output_stream: 'out'
+          output_stream: 'previous2'
+        }
+        node {
+          calculator: 'PacketOnCloseCalculator'
+          input_stream: 'out'
+          output_stream: 'close_out'
+        }
+      )");
+  tool::AddVectorSink("close_out", &graph_config_, &outputs);
+
+  CalculatorGraph graph_;
+  MP_ASSERT_OK(graph_.Initialize(graph_config_, {}));
+  MP_ASSERT_OK(graph_.StartRun({}));
+
+  auto send_packet = [&graph_](const std::string& input_name, int n) {
+    MP_EXPECT_OK(graph_.AddPacketToInputStream(
+        input_name, MakePacket<int>(n).At(Timestamp(n))));
+  };
+
+  send_packet("in", 0);
+  MP_EXPECT_OK(graph_.WaitUntilIdle());
+  EXPECT_EQ(TimestampValues(outputs), (std::vector<int64>{0}));
+
+  for (int main_ts = 1; main_ts < 50; ++main_ts) {
+    send_packet("in", main_ts);
+    MP_EXPECT_OK(graph_.WaitUntilIdle());
+    std::vector<int64> ts_values = TimestampValues(outputs);
+    EXPECT_EQ(ts_values.size(), main_ts);
+    for (int j = 0; j < main_ts; ++j) {
+      CHECK_EQ(ts_values[j], j);
+    }
+  }
+
+  MP_EXPECT_OK(graph_.CloseAllInputStreams());
+  MP_EXPECT_OK(graph_.WaitUntilIdle());
+  MP_EXPECT_OK(graph_.WaitUntilDone());
+}
+
 }  // anonymous namespace
 }  // namespace mediapipe

mediapipe/docs/examples.md

@@ -164,13 +164,13 @@ Below are code samples on how to run MediaPipe on Google Coral Dev Board.
 
 ### Object Detection on Coral
 
-[Object Detection on Coral with Webcam](https://github.com/google/mediapipe/tree/master/mediapipe/examples/coral/README.md)
+[Object Detection on Coral with Webcam](./object_detection_coral_devboard.md)
 shows how to run quantized object detection TFlite model accelerated with
 EdgeTPU on
 [Google Coral Dev Board](https://coral.withgoogle.com/products/dev-board).
 
 ### Face Detection on Coral
 
-[Face Detection on Coral with Webcam](https://github.com/google/mediapipe/tree/master/mediapipe/examples/coral/README.md)
-shows how to use quantized face detection TFlite model accelerated with EdgeTPU
-on [Google Coral Dev Board](https://coral.withgoogle.com/products/dev-board).
+[Face Detection on Coral with Webcam](./face_detection_coral_devboard.md) shows
+how to use quantized face detection TFlite model accelerated with EdgeTPU on
+[Google Coral Dev Board](https://coral.withgoogle.com/products/dev-board).

mediapipe/docs/face_detection_coral_devboard.md

@@ -0,0 +1,20 @@
+## Face Detection on Coral with Webcam
+
+MediaPipe is able to run cross platform across device types like desktop, mobile
+and edge devices. Here is an example of running MediaPipe
+[face detection pipeline](./face_detection_desktop.md) on edge device like
+[Google Coral dev board](https://coral.withgoogle.com/products/dev-board) with
+[Edge TPU](https://cloud.google.com/edge-tpu/). This MediaPipe Coral face
+detection pipeline is running [coral specific quantized version](https://github.com/google/mediapipe/blob/master/mediapipe/examples/coral/models/face-detector-quantized_edgetpu.tflite)
+of the [MediaPipe face detection TFLite model](https://github.com/google/mediapipe/blob/master/mediapipe/models/face_detection_front.tflite)
+accelerated on Edge TPU.
+
+### Cross compilation of MediaPipe Coral binaries in Docker
+
+We recommend building the MediaPipe binaries not on the edge device due to
+limited compute resulting in long build times. Instead, we will build MediaPipe
+binaries using Docker containers on a more powerful host machine. For step by
+step details of cross compiling and running MediaPipe binaries on Coral dev
+board, please refer to [README.md in MediaPipe Coral example folder](https://github.com/google/mediapipe/blob/master/mediapipe/examples/coral/README.md).
+
+![Face Detection running on Coral](images/face_detection_demo_coral.jpg)

mediapipe/docs/object_detection_coral_devboard.md

@@ -0,0 +1,20 @@
+## Object Detection on Coral with Webcam
+
+MediaPipe is able to run cross platform across device types like desktop, mobile
+and edge devices. Here is an example of running MediaPipe
+[object detection pipeline](./object_detection_desktop.md) on edge device like
+[Google Coral dev board](https://coral.withgoogle.com/products/dev-board) with
+[Edge TPU](https://cloud.google.com/edge-tpu/). This MediaPipe Coral object
+detection pipeline is running [coral specific quantized version](https://github.com/google/mediapipe/blob/master/mediapipe/examples/coral/models/object-detector-quantized_edgetpu.tflite)
+of the [MediaPipe object detection TFLite model](https://github.com/google/mediapipe/blob/master/mediapipe/models/object_detection_front.tflite)
+accelerated on Edge TPU.
+
+### Cross compilation of MediaPipe Coral binaries in Docker
+
+We recommend building the MediaPipe binaries not on the edge device due to
+limited compute resulting in long build times. Instead, we will build MediaPipe
+binaries using Docker containers on a more powerful host machine. For step by
+step details of cross compiling and running MediaPipe binaries on Coral dev
+board, please refer to [README.md in MediaPipe Coral example folder](https://github.com/google/mediapipe/blob/master/mediapipe/examples/coral/README.md).
+
+![Object Detection running on Coral](images/object_detection_demo_coral.jpg)