From 54f83e153549dd1a63bcc8fa5e55fb171621a989 Mon Sep 17 00:00:00 2001
From: AlexeyAB <alexeyab84@gmail.com>
Date: Wed, 17 Jan 2018 18:05:07 +0000
Subject: [PATCH] Some fixes

---
 src/yolo_v2_class.hpp |   79 ++++++++++++++++++---------------------
 1 files changed, 36 insertions(+), 43 deletions(-)

diff --git a/src/yolo_v2_class.hpp b/src/yolo_v2_class.hpp
index d60f359..78cc67b 100644
--- a/src/yolo_v2_class.hpp
+++ b/src/yolo_v2_class.hpp
@@ -8,11 +8,6 @@
 #include <opencv2/opencv.hpp>			// C++
 #include "opencv2/highgui/highgui_c.h"	// C
 #include "opencv2/imgproc/imgproc_c.h"	// C
-
-#include <opencv2/cudaoptflow.hpp>
-#include <opencv2/cudaimgproc.hpp>
-#include <opencv2/cudaarithm.hpp>
-#include <opencv2/core/cuda.hpp>
 #endif	// OPENCV
 
 #ifdef YOLODLL_EXPORTS
@@ -47,8 +42,10 @@
 class Detector {
 	std::shared_ptr<void> detector_gpu_ptr;
 	std::deque<std::vector<bbox_t>> prev_bbox_vec_deque;
+	const int cur_gpu_id;
 public:
 	float nms = .4;
+	bool wait_stream;
 
 	YOLODLL_API Detector(std::string cfg_filename, std::string weight_filename, int gpu_id = 0);
 	YOLODLL_API ~Detector();
@@ -89,12 +86,13 @@
 		return mat_to_image(det_mat);
 	}
 
-	static std::shared_ptr<image_t> mat_to_image(cv::Mat img)
+	static std::shared_ptr<image_t> mat_to_image(cv::Mat img_src)
 	{
+		cv::Mat img;
+		cv::cvtColor(img_src, img, cv::COLOR_RGB2BGR);
 		std::shared_ptr<image_t> image_ptr(new image_t, [](image_t *img) { free_image(*img); delete img; });
 		std::shared_ptr<IplImage> ipl_small = std::make_shared<IplImage>(img);
 		*image_ptr = ipl_to_image(ipl_small.get());
-		rgbgr_image(*image_ptr);
 		return image_ptr;
 	}
 
@@ -108,15 +106,17 @@
 		int c = src->nChannels;
 		int step = src->widthStep;
 		image_t out = make_image_custom(w, h, c);
-		int i, j, k, count = 0;;
+		int count = 0;
 
-		for (k = 0; k < c; ++k) {
-			for (i = 0; i < h; ++i) {
-				for (j = 0; j < w; ++j) {
-					out.data[count++] = data[i*step + j*c + k] / 255.;
+		for (int k = 0; k < c; ++k) {
+			for (int i = 0; i < h; ++i) {
+				int i_step = i*step;
+				for (int j = 0; j < w; ++j) {
+					out.data[count++] = data[i_step + j*c + k] / 255.;
 				}
 			}
 		}
+
 		return out;
 	}
 
@@ -137,16 +137,6 @@
 		return out;
 	}
 
-	static void rgbgr_image(image_t im)
-	{
-		int i;
-		for (i = 0; i < im.w*im.h; ++i) {
-			float swap = im.data[i];
-			im.data[i] = im.data[i + im.w*im.h * 2];
-			im.data[i + im.w*im.h * 2] = swap;
-		}
-	}
-
 #endif	// OPENCV
 
 };
@@ -154,24 +144,30 @@
 
 #if defined(TRACK_OPTFLOW) && defined(OPENCV)
 
+#include <opencv2/cudaoptflow.hpp>
+#include <opencv2/cudaimgproc.hpp>
+#include <opencv2/cudaarithm.hpp>
+#include <opencv2/core/cuda.hpp>
+
 class Tracker_optflow {
 public:
-	int gpu_id;
+	const int gpu_count;
+	const int gpu_id;
 
-	Tracker_optflow(int _gpu_id = 0) : gpu_id(_gpu_id)
+
+	Tracker_optflow(int _gpu_id = 0) : gpu_count(cv::cuda::getCudaEnabledDeviceCount()), gpu_id(std::min(_gpu_id, gpu_count-1))
 	{
 		int const old_gpu_id = cv::cuda::getDevice();
-		static const int gpu_count = cv::cuda::getCudaEnabledDeviceCount();
-		if (gpu_count > gpu_id)
-			cv::cuda::setDevice(gpu_id);
+		cv::cuda::setDevice(gpu_id);
+
+		stream = cv::cuda::Stream();
 
 		sync_PyrLKOpticalFlow_gpu = cv::cuda::SparsePyrLKOpticalFlow::create();
-		//sync_PyrLKOpticalFlow_gpu->setWinSize(cv::Size(31, 31)); //sync_PyrLKOpticalFlow_gpu.winSize = cv::Size(31, 31);
-		//sync_PyrLKOpticalFlow_gpu->setWinSize(cv::Size(15, 15)); //sync_PyrLKOpticalFlow_gpu.winSize = cv::Size(15, 15);
+		sync_PyrLKOpticalFlow_gpu->setWinSize(cv::Size(21, 21));	// 15, 21, 31
+		sync_PyrLKOpticalFlow_gpu->setMaxLevel(3);		// +- 3 pt
+		sync_PyrLKOpticalFlow_gpu->setNumIters(2000);	// def: 30
 
-		sync_PyrLKOpticalFlow_gpu->setWinSize(cv::Size(21, 21));
-		sync_PyrLKOpticalFlow_gpu->setMaxLevel(50);	//sync_PyrLKOpticalFlow_gpu.maxLevel = 8;	// +-32 points	// def: 3
-		sync_PyrLKOpticalFlow_gpu->setNumIters(6000);	//sync_PyrLKOpticalFlow_gpu.iters = 8000;	// def: 30
+		cv::cuda::setDevice(old_gpu_id);
 	}
 
 	// just to avoid extra allocations
@@ -183,16 +179,14 @@
 
 	cv::cuda::GpuMat src_grey_gpu;	// used in both functions
 	cv::Ptr<cv::cuda::SparsePyrLKOpticalFlow> sync_PyrLKOpticalFlow_gpu;
+	cv::cuda::Stream stream;
 
 	void update_tracking_flow(cv::Mat src_mat)
 	{
 		int const old_gpu_id = cv::cuda::getDevice();
-		static const int gpu_count = cv::cuda::getCudaEnabledDeviceCount();
-		if (gpu_count > gpu_id)
+		if (old_gpu_id != gpu_id)
 			cv::cuda::setDevice(gpu_id);
 
-		cv::cuda::Stream stream;
-
 		if (src_mat.channels() == 3) {
 			if (src_mat_gpu.cols == 0) {
 				src_mat_gpu = cv::cuda::GpuMat(src_mat.size(), src_mat.type());
@@ -202,7 +196,8 @@
 			src_mat_gpu.upload(src_mat, stream);
 			cv::cuda::cvtColor(src_mat_gpu, src_grey_gpu, CV_BGR2GRAY, 0, stream);
 		}
-		cv::cuda::setDevice(old_gpu_id);
+		if (old_gpu_id != gpu_id)
+			cv::cuda::setDevice(old_gpu_id);
 	}
 
 
@@ -214,21 +209,18 @@
 		}
 
 		int const old_gpu_id = cv::cuda::getDevice();
-		static const int gpu_count = cv::cuda::getCudaEnabledDeviceCount();
-		if (gpu_count > gpu_id)
+		if(old_gpu_id != gpu_id)
 			cv::cuda::setDevice(gpu_id);
 
-		cv::cuda::Stream stream;
-
 		if (dst_mat_gpu.cols == 0) {
 			dst_mat_gpu = cv::cuda::GpuMat(dst_mat.size(), dst_mat.type());
 			dst_grey_gpu = cv::cuda::GpuMat(dst_mat.size(), CV_8UC1);
 			tmp_grey_gpu = cv::cuda::GpuMat(dst_mat.size(), CV_8UC1);
 		}
 
-
 		dst_mat_gpu.upload(dst_mat, stream);
 
+
 		cv::cuda::cvtColor(dst_mat_gpu, dst_grey_gpu, CV_BGR2GRAY, 0, stream);
 
 		if (src_grey_gpu.rows != dst_grey_gpu.rows || src_grey_gpu.cols != dst_grey_gpu.cols) {
@@ -299,7 +291,8 @@
 					}
 		}
 
-		cv::cuda::setDevice(old_gpu_id);
+		if (old_gpu_id != gpu_id)
+			cv::cuda::setDevice(old_gpu_id);
 
 		return result_bbox_vec;
 	}

--
Gitblit v1.10.0