Build category mask without resizing tensors to image size.

PiperOrigin-RevId: 523754567
This commit is contained in:
MediaPipe Team 2023-04-12 11:19:40 -07:00 committed by Copybara-Service
parent 049ba8bbca
commit 776dceb588

View File

@ -14,6 +14,7 @@ limitations under the License.
==============================================================================*/ ==============================================================================*/
#include <algorithm> #include <algorithm>
#include <cmath>
#include <cstdint> #include <cstdint>
#include <limits> #include <limits>
#include <memory> #include <memory>
@ -81,25 +82,34 @@ void Sigmoid(absl::Span<const float> values,
[](float value) { return 1. / (1 + std::exp(-value)); }); [](float value) { return 1. / (1 + std::exp(-value)); });
} }
// Linearly interpolate the value between v0 and v1. Assume 0 <= t <= 1.
float LinearInterpolate(float v0, float v1, float t) {
DCHECK(t >= 0 && t <= 1);
return v0 + (v1 - v0) * t;
}
// Bilinearly interpolate the value between 4 points. Assume 0 <= t0, t1 <= 1.
float BilinearInterpolate(float v00, float v10, float v01, float v11, float t0,
float t1) {
DCHECK(t0 >= 0 && t0 <= 1 && t1 >= 0 && t1 <= 1);
return LinearInterpolate(LinearInterpolate(v00, v10, t0),
LinearInterpolate(v01, v11, t0), t1);
}
float GetTensorElement(const Shape& input_shape, const float* tensors_buffer,
int x, int y, int c) {
return tensors_buffer[y * input_shape.channels * input_shape.width +
x * input_shape.channels + c];
}
Image ProcessForCategoryMaskCpu(const Shape& input_shape, Image ProcessForCategoryMaskCpu(const Shape& input_shape,
const Shape& output_shape, const Shape& output_shape,
const SegmenterOptions& options, const SegmenterOptions& options,
const float* tensors_buffer) { const float* tensors_buffer) {
cv::Mat resized_tensors_mat; const float width_scale =
cv::Mat tensors_mat_view( (input_shape.width - 1) / static_cast<float>(output_shape.width - 1);
input_shape.height, input_shape.width, CV_32FC(input_shape.channels), const float height_scale =
reinterpret_cast<void*>(const_cast<float*>(tensors_buffer))); (input_shape.height - 1) / static_cast<float>(output_shape.height - 1);
if (output_shape.height == input_shape.height &&
output_shape.width == input_shape.width) {
resized_tensors_mat = tensors_mat_view;
} else {
// Resize input tensors to output size.
// TOOD(b/273633027) Use an efficient way to find values for category mask
// instead of resizing the whole tensor .
cv::resize(tensors_mat_view, resized_tensors_mat,
{output_shape.width, output_shape.height}, 0, 0,
cv::INTER_LINEAR);
}
// Category mask Image. // Category mask Image.
ImageFrameSharedPtr image_frame_ptr = std::make_shared<ImageFrame>( ImageFrameSharedPtr image_frame_ptr = std::make_shared<ImageFrame>(
@ -111,23 +121,37 @@ Image ProcessForCategoryMaskCpu(const Shape& input_shape,
mediapipe::formats::MatView(image_frame_ptr.get()); mediapipe::formats::MatView(image_frame_ptr.get());
const int input_channels = input_shape.channels; const int input_channels = input_shape.channels;
category_mask_mat_view.forEach<uint8_t>( category_mask_mat_view.forEach<uint8_t>(
[&resized_tensors_mat, &input_channels, &options](uint8_t& pixel, [&tensors_buffer, &input_shape, &width_scale, &height_scale,
const int position[]) { &input_channels, &options](uint8_t& pixel, const int position[]) {
float* tensors_buffer = std::vector<float> confidence_scores(input_channels);
resized_tensors_mat.ptr<float>(position[0], position[1]); int y0 = static_cast<int>(std::floor(position[0] * height_scale));
absl::Span<float> confidence_scores(tensors_buffer, input_channels); int x0 = static_cast<int>(std::floor(position[1] * width_scale));
int y1 = static_cast<int>(std::ceil(position[0] * height_scale));
int x1 = static_cast<int>(std::ceil(position[1] * width_scale));
float t0 = position[0] * height_scale - y0;
float t1 = position[1] * width_scale - x0;
for (int i = 0; i < input_channels; ++i) {
confidence_scores[i] = BilinearInterpolate(
GetTensorElement(input_shape, tensors_buffer, x0, y0, i),
GetTensorElement(input_shape, tensors_buffer, x0, y1, i),
GetTensorElement(input_shape, tensors_buffer, x1, y0, i),
GetTensorElement(input_shape, tensors_buffer, x1, y1, i), t0, t1);
}
absl::Span<float> confidence_scores_span(confidence_scores.data(),
confidence_scores.size());
// Only process the activation function if it is SIGMOID. If NONE, // Only process the activation function if it is SIGMOID. If NONE,
// we do nothing for activation, If SOFTMAX, it is required // we do nothing for activation, If SOFTMAX, it is required
// to have input_channels > 1, and for input_channels > 1, we don't need // to have input_channels > 1, and for input_channels > 1, we don't need
// activation to find the maximum value. // activation to find the maximum value.
if (options.activation() == SegmenterOptions::SIGMOID) { if (options.activation() == SegmenterOptions::SIGMOID) {
Sigmoid(confidence_scores, confidence_scores); Sigmoid(confidence_scores_span, confidence_scores_span);
} }
if (input_channels == 1) { if (input_channels == 1) {
// if the input tensor is a single mask, it is assumed to be a binary // if the input tensor is a single mask, it is assumed to be a binary
// foreground segmentation mask. For such a mask, we make foreground // foreground segmentation mask. For such a mask, we make foreground
// category 1, and background category 0. // category 1, and background category 0.
pixel = static_cast<uint8_t>(*tensors_buffer > 0.5f); pixel = static_cast<uint8_t>(confidence_scores[0] > 0.5f);
} else { } else {
const int maximum_category_idx = const int maximum_category_idx =
std::max_element(confidence_scores.begin(), std::max_element(confidence_scores.begin(),