Merge branch 'main' into nvtransformchange

.github/workflows/build-test-lin-container.yml

@@ -104,7 +104,7 @@ jobs:
         submodules: 'recursive' 
         lfs: true
         fetch-depth: 0
-
+        
     - name: List Submodules
       run: |
         git config --global --add safe.directory "*"

base/CMakeLists.txt

@@ -259,6 +259,7 @@ SET(IP_FILES
 	src/ImageViewerModule.cpp
 	src/BMPConverter.cpp
 	src/ImageResizeCV.cpp
+	src/FacialLandmarksCV.cpp
 	src/ImageEncoderCV.cpp
 	src/RotateCV.cpp
 	src/BrightnessContrastControlXform.cpp
@@ -288,6 +289,7 @@ SET(IP_FILES_H
 	include/ImageDecoderCV.h
 	include/BMPConverter.h
 	include/ImageResizeCV.h
+	include/FacialLandmarksCV.h
 	include/ImageEncoderCV.h
 	include/RotateCV.h
 	include/BrightnessContrastControlXform.h
@@ -478,6 +480,7 @@ IF (ENABLE_CUDA)
 		test/cudamemcopy_tests.cpp		
 		test/resizenppi_tests.cpp
 		test/rotatenppi_tests.cpp
+		test/ccnppi_tests.cpp
 	)
 	IF(NOT ENABLE_ARM64) # following tests need CUDA but can not run on ARM ?
 
@@ -486,7 +489,6 @@ IF (ENABLE_CUDA)
 			test/jpegdecodernvjpeg_tests.cpp			
 			test/resizenppi_jpegencodernvjpeg_tests.cpp
 			test/nvjpeg_combo_tests.cpp			
-			test/ccnppi_tests.cpp
 			test/overlaynppi_tests.cpp
 			test/effectsnppi_tests.cpp
 			test/h264Encodernvcodec_tests.cpp
@@ -520,6 +522,7 @@ SET(UT_FILES
 	test/findexstrategy_tests.cpp	
 	test/jpegdecodercv_tests.cpp
 	test/Imageresizecv_tests.cpp
+	test/faciallandmarkscv_tests.cpp
 	test/ImageEncodeCV_tests.cpp
 	test/rotatecv_tests.cpp
 	test/brightness_contrast_tests.cpp

base/include/ApraPoint2f.h

@@ -23,6 +23,11 @@ class ApraPoint2f : public cv::Point2f
 
 	}
 
+	ApraPoint2f(cv::Point2f &point) : cv::Point2f(point)
+	{
+
+	}
+
 private:
 	friend class boost::serialization::access;
 

base/include/CCNPPI.h

@@ -8,6 +8,7 @@
 class CCNPPIProps : public ModuleProps
 {
 public:
+
 	CCNPPIProps(ImageMetadata::ImageType _imageType, cudastream_sp& _stream)
 	{
 		stream_sp = _stream;
@@ -49,6 +50,7 @@ class CCNPPI : public Module
 	int mOutputFrameType;
 	size_t mFrameLength;
 	framemetadata_sp mOutputMetadata;
+	framemetadata_sp mIntermediateMetadata;
 	std::string mOutputPinId;
-	CCNPPIProps props;		
+	CCNPPIProps mProps;		
 };

base/include/FacialLandmarksCV.h

@@ -0,0 +1,84 @@
+#pragma once
+
+#include <opencv2/face.hpp>
+#include "Module.h"
+
+class Detail;
+class DetailSSD;
+class DetailHCASCADE;
+
+class FacialLandmarkCVProps : public ModuleProps
+{
+public:
+	enum FaceDetectionModelType
+	{
+		SSD,
+		HAAR_CASCADE
+	};
+
+	FacialLandmarkCVProps(FaceDetectionModelType _type) : type(_type) {}
+
+	FacialLandmarkCVProps(FaceDetectionModelType _type, const std::string _Face_Detection_Configuration, const std::string _Face_Detection_Weights, const std::string _landmarksDetectionModel, cv::Ptr<cv::face::Facemark> _facemark)
+		: type(_type), Face_Detection_Configuration(_Face_Detection_Configuration), Face_Detection_Weights(_Face_Detection_Weights), landmarksDetectionModel(_landmarksDetectionModel),facemark(_facemark)
+	{
+		if (_type != FaceDetectionModelType::SSD)
+		{
+			 throw AIPException(AIP_FATAL, "This constructor only supports SSD");
+		}
+	}
+
+	FacialLandmarkCVProps(FaceDetectionModelType _type, const std::string _faceDetectionModel,const std::string _landmarksDetectionModel, cv::Ptr<cv::face::Facemark> _facemark)
+		: type(_type), landmarksDetectionModel(_landmarksDetectionModel), faceDetectionModel(_faceDetectionModel), facemark(_facemark)
+	{
+		if (_type != FaceDetectionModelType::HAAR_CASCADE)
+		{
+			throw AIPException(AIP_FATAL, "This constructor only supports HAAR_CASCADE ");
+		}
+	}
+
+	FaceDetectionModelType type;
+	const std::string Face_Detection_Configuration = "./data/assets/deploy.prototxt";
+	const std::string Face_Detection_Weights = "./data/assets/res10_300x300_ssd_iter_140000_fp16.caffemodel";
+	const std::string landmarksDetectionModel = "./data/assets/face_landmark_model.dat";
+	const std::string faceDetectionModel = "./data/assets/haarcascade.xml";
+	cv::Ptr<cv::face::Facemark> facemark = cv::face::FacemarkKazemi::create();
+
+	size_t getSerializeSize()
+	{
+		return ModuleProps::getSerializeSize() + sizeof(type);
+	}
+
+private:
+	friend class boost::serialization::access;
+
+	template <class Archive>
+	void serialize(Archive& ar, const unsigned int version)
+	{
+		ar& boost::serialization::base_object<ModuleProps>(*this);
+		ar& type;
+	}
+};
+
+class FacialLandmarkCV : public Module
+{
+ public:
+	FacialLandmarkCV(FacialLandmarkCVProps props);
+	virtual ~FacialLandmarkCV();
+	bool init();
+	bool term();
+	void setProps(FacialLandmarkCVProps& props);
+	FacialLandmarkCVProps getProps();
+
+protected:
+	bool process(frame_container &frames);
+	bool processSOS(frame_sp &frame);
+	bool validateInputPins();
+	bool validateOutputPins();
+	void addInputPin(framemetadata_sp &metadata, string &pinId); // throws exception if validation fails
+	bool shouldTriggerSOS();
+	bool processEOS(string &pinId);
+	boost::shared_ptr<Detail> mDetail;
+	FacialLandmarkCVProps mProp;
+	bool handlePropsChange(frame_sp& frame);
+	std::string mOutputPinId1;
+};

base/include/FrameMetadata.h

@@ -49,7 +49,8 @@ class FrameMetadata {
 		MP4_VIDEO_METADATA,
 		HEVC_DATA, //H265
 		MOTION_VECTOR_DATA,
-		OVERLAY_INFO_IMAGE
+		OVERLAY_INFO_IMAGE,
+		FACE_LANDMARKS_INFO
 	};
 
 	enum MemType

base/include/Mp4WriterSink.h

@@ -9,11 +9,12 @@ class DetailH264;
 class Mp4WriterSinkProps : public ModuleProps
 {
 public:
-	Mp4WriterSinkProps(uint32_t _chunkTime, uint32_t _syncTimeInSecs, uint16_t _fps, std::string _baseFolder, bool _recordedTSBasedDTS = true) : ModuleProps()
+	Mp4WriterSinkProps(uint32_t _chunkTime, uint32_t _syncTimeInSecs, uint16_t _fps, std::string _baseFolder, bool _recordedTSBasedDTS = true,  bool _enableMetadata = true) : ModuleProps()
 	{
 		baseFolder = _baseFolder;
 		fps = _fps;
 		recordedTSBasedDTS = _recordedTSBasedDTS;
+		enableMetadata = _enableMetadata;
 		if ((_chunkTime >= 1 && _chunkTime <= 60) || (_chunkTime == UINT32_MAX))
 		{
 			chunkTime = _chunkTime;
@@ -39,6 +40,7 @@ class Mp4WriterSinkProps : public ModuleProps
 		syncTimeInSecs = 1;
 		fps = 30;
 		recordedTSBasedDTS = true;
+		enableMetadata = true;
 	}
 
 	size_t getSerializeSize()
@@ -48,14 +50,16 @@ class Mp4WriterSinkProps : public ModuleProps
 			sizeof(baseFolder) +
 			sizeof(chunkTime) +
 			sizeof(syncTimeInSecs) +
-			sizeof(fps);
+			sizeof(fps) +
+			sizeof(enableMetadata);;
 	}
 
 	std::string baseFolder;
 	uint32_t chunkTime = 1;
 	uint32_t syncTimeInSecs = 1;
 	uint16_t fps = 30;
 	bool recordedTSBasedDTS = true;
+	bool enableMetadata = true;
 private:
 	friend class boost::serialization::access;
 
@@ -68,6 +72,7 @@ class Mp4WriterSinkProps : public ModuleProps
 		ar &chunkTime;
 		ar &syncTimeInSecs;
 		ar &fps;
+		ar &enableMetadata;
 	}
 };
 
@@ -89,6 +94,8 @@ class Mp4WriterSink : public Module
 	bool setMetadata(framemetadata_sp &inputMetadata);
 	bool handlePropsChange(frame_sp &frame);
 	bool shouldTriggerSOS();
+	void addInputPin(framemetadata_sp& metadata, string& pinId);
+	bool enableMp4Metadata(framemetadata_sp &inputMetadata);
 	boost::shared_ptr<DetailAbs> mDetail;
 	Mp4WriterSinkProps mProp;
 

base/include/Mp4WriterSinkUtils.h

@@ -13,7 +13,7 @@ class Mp4WriterSinkUtils
 	bool customNamedFileDirCheck(std::string baseFolder, uint32_t chunkTimeInMinutes, boost::filesystem::path relPath, std::string& nextFrameFileName);
 	std::string format_hrs(int &hr);
 	std::string format_2(int &min);
-	std::string filePath(boost::filesystem::path relPath, std::string mp4FileName, std::string baseFolder);
+	std::string filePath(boost::filesystem::path relPath, std::string mp4FileName, std::string baseFolder, uint64_t chunkTimeInMins);
 	~Mp4WriterSinkUtils();
 private:
 	int lastVideoMinute=0;