From a2f3644e87c11c222f86aa2aeb0e31d858ebb024 Mon Sep 17 00:00:00 2001
From: Alexey <AlexeyAB@users.noreply.github.com>
Date: Sat, 21 Apr 2018 20:56:46 +0000
Subject: [PATCH] Merge pull request #675 from IlyaOvodov/Flip_Fix
---
src/yolo_v2_class.cpp | 43 +++++++++++++++++++++----------------------
1 files changed, 21 insertions(+), 22 deletions(-)
diff --git a/src/yolo_v2_class.cpp b/src/yolo_v2_class.cpp
index c805e29..03d581a 100644
--- a/src/yolo_v2_class.cpp
+++ b/src/yolo_v2_class.cpp
@@ -22,16 +22,16 @@
#define FRAMES 3
+#ifdef GPU
void check_cuda(cudaError_t status) {
if (status != cudaSuccess) {
const char *s = cudaGetErrorString(status);
printf("CUDA Error Prev: %s\n", s);
}
}
+#endif
struct detector_gpu_t {
- float **probs;
- box *boxes;
network net;
image images[FRAMES];
float *avg;
@@ -52,11 +52,12 @@
detector_gpu_t &detector_gpu = *static_cast<detector_gpu_t *>(detector_gpu_ptr.get());
#ifdef GPU
- check_cuda( cudaSetDevice(gpu_id) );
- printf(" Used GPU %d \n", gpu_id);
+ //check_cuda( cudaSetDevice(cur_gpu_id) );
+ cuda_set_device(cur_gpu_id);
+ printf(" Used GPU %d \n", cur_gpu_id);
#endif
network &net = detector_gpu.net;
- net.gpu_index = gpu_id;
+ net.gpu_index = cur_gpu_id;
//gpu_index = i;
char *cfgfile = const_cast<char *>(cfg_filename.data());
@@ -67,7 +68,8 @@
load_weights(&net, weightfile);
}
set_batch_network(&net, 1);
- net.gpu_index = gpu_id;
+ net.gpu_index = cur_gpu_id;
+ fuse_conv_batchnorm(net);
layer l = net.layers[net.n - 1];
int j;
@@ -76,10 +78,6 @@
for (j = 0; j < FRAMES; ++j) detector_gpu.predictions[j] = (float *)calloc(l.outputs, sizeof(float));
for (j = 0; j < FRAMES; ++j) detector_gpu.images[j] = make_image(1, 1, 3);
- detector_gpu.boxes = (box *)calloc(l.w*l.h*l.n, sizeof(box));
- detector_gpu.probs = (float **)calloc(l.w*l.h*l.n, sizeof(float *));
- for (j = 0; j < l.w*l.h*l.n; ++j) detector_gpu.probs[j] = (float *)calloc(l.classes, sizeof(float));
-
detector_gpu.track_id = (unsigned int *)calloc(l.classes, sizeof(unsigned int));
for (j = 0; j < l.classes; ++j) detector_gpu.track_id[j] = 1;
@@ -100,14 +98,10 @@
for (int j = 0; j < FRAMES; ++j) free(detector_gpu.predictions[j]);
for (int j = 0; j < FRAMES; ++j) if(detector_gpu.images[j].data) free(detector_gpu.images[j].data);
- for (int j = 0; j < l.w*l.h*l.n; ++j) free(detector_gpu.probs[j]);
- free(detector_gpu.boxes);
- free(detector_gpu.probs);
-
int old_gpu_index;
#ifdef GPU
cudaGetDevice(&old_gpu_index);
- cudaSetDevice(detector_gpu.net.gpu_index);
+ cuda_set_device(detector_gpu.net.gpu_index);
#endif
free_network(detector_gpu.net);
@@ -221,17 +215,21 @@
l.output = detector_gpu.avg;
detector_gpu.demo_index = (detector_gpu.demo_index + 1) % FRAMES;
}
+ //get_region_boxes(l, 1, 1, thresh, detector_gpu.probs, detector_gpu.boxes, 0, 0);
+ //if (nms) do_nms_sort(detector_gpu.boxes, detector_gpu.probs, l.w*l.h*l.n, l.classes, nms);
- get_region_boxes(l, 1, 1, thresh, detector_gpu.probs, detector_gpu.boxes, 0, 0);
- if (nms) do_nms_sort(detector_gpu.boxes, detector_gpu.probs, l.w*l.h*l.n, l.classes, nms);
- //draw_detections(im, l.w*l.h*l.n, thresh, boxes, probs, names, alphabet, l.classes);
+ int nboxes = 0;
+ int letterbox = 0;
+ float hier_thresh = 0.5;
+ detection *dets = get_network_boxes(&net, im.w, im.h, thresh, hier_thresh, 0, 1, &nboxes, letterbox);
+ if (nms) do_nms_sort(dets, nboxes, l.classes, nms);
std::vector<bbox_t> bbox_vec;
- for (size_t i = 0; i < (l.w*l.h*l.n); ++i) {
- box b = detector_gpu.boxes[i];
- int const obj_id = max_index(detector_gpu.probs[i], l.classes);
- float const prob = detector_gpu.probs[i][obj_id];
+ for (size_t i = 0; i < nboxes; ++i) {
+ box b = dets[i].bbox;
+ int const obj_id = max_index(dets[i].prob, l.classes);
+ float const prob = dets[i].prob[obj_id];
if (prob > thresh)
{
@@ -248,6 +246,7 @@
}
}
+ free_detections(dets, nboxes);
if(sized.data)
free(sized.data);
--
Gitblit v1.10.0