talksik/mediapipe
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Project import generated by Copybara.

Browse Source 
GitOrigin-RevId: afeb9cf5a8c069c0a566d16e1622bbb086170e4d
This commit is contained in:
MediaPipe Team
2020-05-21 13:37:51 -04:00
committed by chuoling
parent b6e680647c
commit b133b0f200
258 changed files with 4146 additions and 5147 deletions
Download Patch File Download Diff File Expand all files Collapse all files
mediapipe/calculators/image/color_convert_calculator.cc
+24
View File
@@ -38,9 +38,11 @@ void SetColorChannel(int channel, uint8 value, cv::Mat* mat) {
constexpr char kRgbaInTag[] = "RGBA_IN";
constexpr char kRgbInTag[] = "RGB_IN";
constexpr char kBgraInTag[] = "BGRA_IN";
constexpr char kGrayInTag[] = "GRAY_IN";
constexpr char kRgbaOutTag[] = "RGBA_OUT";
constexpr char kRgbOutTag[] = "RGB_OUT";
constexpr char kBgraOutTag[] = "BGRA_OUT";
constexpr char kGrayOutTag[] = "GRAY_OUT";
} // namespace
@@ -53,6 +55,8 @@ constexpr char kGrayOutTag[] = "GRAY_OUT";
// GRAY -> RGB
// RGB -> GRAY
// RGB -> RGBA
// RGBA -> BGRA
// BGRA -> RGBA
//
// This calculator only supports a single input stream and output stream at a
// time. If more than one input stream or output stream is present, the
@@ -63,11 +67,13 @@ constexpr char kGrayOutTag[] = "GRAY_OUT";
// Input streams:
// RGBA_IN: The input video stream (ImageFrame, SRGBA).
// RGB_IN: The input video stream (ImageFrame, SRGB).
// BGRA_IN: The input video stream (ImageFrame, SBGRA).
// GRAY_IN: The input video stream (ImageFrame, GRAY8).
//
// Output streams:
// RGBA_OUT: The output video stream (ImageFrame, SRGBA).
// RGB_OUT: The output video stream (ImageFrame, SRGB).
// BGRA_OUT: The output video stream (ImageFrame, SBGRA).
// GRAY_OUT: The output video stream (ImageFrame, GRAY8).
class ColorConvertCalculator : public CalculatorBase {
public:
@@ -113,6 +119,10 @@ REGISTER_CALCULATOR(ColorConvertCalculator);
cc->Inputs().Tag(kRgbInTag).Set<ImageFrame>();
}
if (cc->Inputs().HasTag(kBgraInTag)) {
cc->Inputs().Tag(kBgraInTag).Set<ImageFrame>();
}
if (cc->Outputs().HasTag(kRgbOutTag)) {
cc->Outputs().Tag(kRgbOutTag).Set<ImageFrame>();
}
@@ -125,6 +135,10 @@ REGISTER_CALCULATOR(ColorConvertCalculator);
cc->Outputs().Tag(kRgbaOutTag).Set<ImageFrame>();
}
if (cc->Outputs().HasTag(kBgraOutTag)) {
cc->Outputs().Tag(kBgraOutTag).Set<ImageFrame>();
}
return ::mediapipe::OkStatus();
}
@@ -171,6 +185,16 @@ REGISTER_CALCULATOR(ColorConvertCalculator);
return ConvertAndOutput(kRgbInTag, kRgbaOutTag, ImageFormat::SRGBA,
cv::COLOR_RGB2RGBA, cc);
}
// BGRA -> RGBA
if (cc->Inputs().HasTag(kBgraInTag) && cc->Outputs().HasTag(kRgbaOutTag)) {
return ConvertAndOutput(kBgraInTag, kRgbaOutTag, ImageFormat::SRGBA,
cv::COLOR_BGRA2RGBA, cc);
}
// RGBA -> BGRA
if (cc->Inputs().HasTag(kRgbaInTag) && cc->Outputs().HasTag(kBgraOutTag)) {
return ConvertAndOutput(kRgbaInTag, kBgraOutTag, ImageFormat::SBGRA,
cv::COLOR_RGBA2BGRA, cc);
}
return ::mediapipe::InvalidArgumentErrorBuilder(MEDIAPIPE_LOC)
<< "Unsupported image format conversion.";
mediapipe/calculators/image/image_cropping_calculator.cc
+1 -7
View File
@@ -514,13 +514,7 @@ RectSpec ImageCroppingCalculator::GetCropSpecs(const CalculatorContext* cc,
}
}
return {
.width = crop_width,
.height = crop_height,
.center_x = x_center,
.center_y = y_center,
.rotation = rotation,
};
return {crop_width, crop_height, x_center, y_center, rotation};
}
::mediapipe::Status ImageCroppingCalculator::GetBorderModeForOpenCV(
mediapipe/calculators/image/image_transformation_calculator.cc
+14 -10
View File
@@ -392,19 +392,26 @@ REGISTER_CALCULATOR(ImageTransformationCalculator);
}
cv::Mat scaled_mat;
int output_width = output_width_;
int output_height = output_height_;
if (scale_mode_ == mediapipe::ScaleMode_Mode_STRETCH) {
cv::resize(input_mat, scaled_mat, cv::Size(output_width_, output_height_));
int scale_flag =
input_mat.cols > output_width_ && input_mat.rows > output_height_
? cv::INTER_AREA
: cv::INTER_LINEAR;
cv::resize(input_mat, scaled_mat, cv::Size(output_width_, output_height_),
0, 0, scale_flag);
} else {
const float scale =
std::min(static_cast<float>(output_width_) / input_width,
static_cast<float>(output_height_) / input_height);
const int target_width = std::round(input_width * scale);
const int target_height = std::round(input_height * scale);
int scale_flag = scale < 1.0f ? cv::INTER_AREA : cv::INTER_LINEAR;
if (scale_mode_ == mediapipe::ScaleMode_Mode_FIT) {
cv::Mat intermediate_mat;
cv::resize(input_mat, intermediate_mat,
cv::Size(target_width, target_height));
cv::Size(target_width, target_height), 0, 0, scale_flag);
const int top = (output_height_ - target_height) / 2;
const int bottom = output_height_ - target_height - top;
const int left = (output_width_ - target_width) / 2;
@@ -413,16 +420,13 @@ REGISTER_CALCULATOR(ImageTransformationCalculator);
options_.constant_padding() ? cv::BORDER_CONSTANT
: cv::BORDER_REPLICATE);
} else {
cv::resize(input_mat, scaled_mat, cv::Size(target_width, target_height));
output_width_ = target_width;
output_height_ = target_height;
cv::resize(input_mat, scaled_mat, cv::Size(target_width, target_height),
0, 0, scale_flag);
output_width = target_width;
output_height = target_height;
}
}
int output_width;
int output_height;
ComputeOutputDimensions(input_width, input_height, &output_width,
&output_height);
if (cc->Outputs().HasTag("LETTERBOX_PADDING")) {
auto padding = absl::make_unique<std::array<float, 4>>();
ComputeOutputLetterboxPadding(input_width, input_height, output_width,