github
/
darknet
mirror of https://github.com/AlexeyAB/darknet.git
1
0
Fork 0
Code Issues Releases Wiki Activity

Merge pull request #741 from IlyaOvodov/Fix_detector_output

Browse Source 
Output improvements for detector results:
pull/844/head
Alexey 7 years ago committed by GitHub
parent 0045074f56 1b5afb4583
commit 0948df52b8
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
6 changed files with 104 additions and 31 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 11
      src/box.h
  2. 2
      src/convolutional_layer.c
  3. 5
      src/darknet.c
  4. 10
      src/detector.c
  5. 105
      src/image.c
  6. 2
      src/image.h

11
src/box.h
Unescape View File

@ -37,6 +37,13 @@ typedef struct detection {
int sort_class;
} detection;
typedef struct detection_with_class {
detection det;
// The most probable class id: the best class index in this->prob.
// Is filled temporary when processing results, otherwise not initialized
int best_class;
} detection_with_class;
box float_to_box(float *f);
float box_iou(box a, box b);
float box_rmse(box a, box b);
@ -48,4 +55,8 @@ YOLODLL_API void do_nms_obj(detection *dets, int total, int classes, float thres
box decode_box(box b, box anchor);
box encode_box(box b, box anchor);
// Creates array of detections with prob > thresh and fills best_class for them
// Return number of selected detections in *selected_detections_num
detection_with_class* get_actual_detections(detection *dets, int dets_num, float thresh, int* selected_detections_num);
#endif

2
src/convolutional_layer.c
Unescape View File

@ -491,7 +491,7 @@ void resize_convolutional_layer(convolutional_layer *l, int w, int h)
size_t total_byte;
check_error(cudaMemGetInfo(&free_byte, &total_byte));
if (l->workspace_size > free_byte || l->workspace_size >= total_byte / 2) {
printf(" used slow CUDNN algo without Workspace! \n");
printf(" used slow CUDNN algo without Workspace! Need memory: %d, available: %d\n", l->workspace_size, (free_byte < total_byte/2) ? free_byte : total_byte/2);
cudnn_convolutional_setup(l, cudnn_smallest);
l->workspace_size = get_workspace_size(*l);
}

5
src/darknet.c
Unescape View File

@ -18,7 +18,7 @@
#endif
extern void predict_classifier(char *datacfg, char *cfgfile, char *weightfile, char *filename, int top);
extern void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh);
extern void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh, int ext_output);
extern void run_voxel(int argc, char **argv);
extern void run_yolo(int argc, char **argv);
extern void run_detector(int argc, char **argv);
@ -391,8 +391,9 @@ int main(int argc, char **argv)
run_detector(argc, argv);
} else if (0 == strcmp(argv[1], "detect")){
float thresh = find_float_arg(argc, argv, "-thresh", .24);
int ext_output = find_arg(argc, argv, "-ext_output");
char *filename = (argc > 4) ? argv[4]: 0;
test_detector("cfg/coco.data", argv[2], argv[3], filename, thresh);
test_detector("cfg/coco.data", argv[2], argv[3], filename, thresh, ext_output);
} else if (0 == strcmp(argv[1], "cifar")){
run_cifar(argc, argv);
} else if (0 == strcmp(argv[1], "go")){

10
src/detector.c
Unescape View File

@ -1041,7 +1041,8 @@ void calc_anchors(char *datacfg, int num_of_clusters, int width, int height, int
}
#endif // OPENCV
void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh, float hier_thresh, int dont_show)
void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh,
float hier_thresh, int dont_show, int ext_output)
{
list *options = read_data_cfg(datacfg);
char *name_list = option_find_str(options, "names", "data/names.list");
@ -1093,7 +1094,7 @@ void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filenam
int nboxes = 0;
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);
draw_detections_v3(im, dets, nboxes, thresh, names, alphabet, l.classes);
draw_detections_v3(im, dets, nboxes, thresh, names, alphabet, l.classes, ext_output);
free_detections(dets, nboxes);
save_image(im, "predictions");
if (!dont_show) {
@ -1145,6 +1146,9 @@ void run_detector(int argc, char **argv)
int num_of_clusters = find_int_arg(argc, argv, "-num_of_clusters", 5);
int width = find_int_arg(argc, argv, "-width", -1);
int height = find_int_arg(argc, argv, "-height", -1);
// extended output in test mode (output of rect bound coords)
// and for recall mode (extended output table-like format with results for best_class fit)
int ext_output = find_arg(argc, argv, "-ext_output");
if(argc < 4){
fprintf(stderr, "usage: %s %s [train/test/valid] [cfg] [weights (optional)]\n", argv[0], argv[1]);
return;
@ -1181,7 +1185,7 @@ void run_detector(int argc, char **argv)
if(strlen(weights) > 0)
if (weights[strlen(weights) - 1] == 0x0d) weights[strlen(weights) - 1] = 0;
char *filename = (argc > 6) ? argv[6]: 0;
if(0==strcmp(argv[2], "test")) test_detector(datacfg, cfg, weights, filename, thresh, hier_thresh, dont_show);
if(0==strcmp(argv[2], "test")) test_detector(datacfg, cfg, weights, filename, thresh, hier_thresh, dont_show, ext_output);
else if(0==strcmp(argv[2], "train")) train_detector(datacfg, cfg, weights, gpus, ngpus, clear, dont_show);
else if(0==strcmp(argv[2], "valid")) validate_detector(datacfg, cfg, weights, outfile);
else if(0==strcmp(argv[2], "recall")) validate_detector_recall(datacfg, cfg, weights);

105
src/image.c
Unescape View File

@ -230,27 +230,76 @@ image **load_alphabet()
return alphabets;
}
void draw_detections_v3(image im, detection *dets, int num, float thresh, char **names, image **alphabet, int classes)
{
int i, j;
for (i = 0; i < num; ++i) {
char labelstr[4096] = { 0 };
int class_id = -1;
for (j = 0; j < classes; ++j) {
if (dets[i].prob[j] > thresh) {
if (class_id < 0) {
strcat(labelstr, names[j]);
class_id = j;
}
else {
strcat(labelstr, ", ");
strcat(labelstr, names[j]);
}
printf("%s: %.0f%%\n", names[j], dets[i].prob[j] * 100);
// Creates array of detections with prob > thresh and fills best_class for them
detection_with_class* get_actual_detections(detection *dets, int dets_num, float thresh, int* selected_detections_num)
{
int selected_num = 0;
detection_with_class* result_arr = calloc(dets_num, sizeof(detection_with_class));
for (int i = 0; i < dets_num; ++i) {
int best_class = -1;
float best_class_prob = thresh;
for (int j = 0; j < dets[i].classes; ++j) {
if (dets[i].prob[j] > best_class_prob ) {
best_class = j;
best_class_prob = dets[i].prob[j];
}
}
if (class_id >= 0) {
if (best_class >= 0) {
result_arr[selected_num].det = dets[i];
result_arr[selected_num].best_class = best_class;
++selected_num;
}
}
if (selected_detections_num)
*selected_detections_num = selected_num;
return result_arr;
}
// compare to sort detection** by bbox.x
int compare_by_lefts(const void *a_ptr, const void *b_ptr) {
const detection_with_class* a = (detection_with_class*)a_ptr;
const detection_with_class* b = (detection_with_class*)b_ptr;
const float delta = (a->det.bbox.x - a->det.bbox.w/2) - (b->det.bbox.x - b->det.bbox.w/2);
return delta < 0 ? -1 : delta > 0 ? 1 : 0;
}
// compare to sort detection** by best_class probability
int compare_by_probs(const void *a_ptr, const void *b_ptr) {
const detection_with_class* a = (detection_with_class*)a_ptr;
const detection_with_class* b = (detection_with_class*)b_ptr;
float delta = a->det.prob[a->best_class] - b->det.prob[b->best_class];
return delta < 0 ? -1 : delta > 0 ? 1 : 0;
}
void draw_detections_v3(image im, detection *dets, int num, float thresh, char **names, image **alphabet, int classes, int ext_output)
{
int selected_detections_num;
detection_with_class* selected_detections = get_actual_detections(dets, num, thresh, &selected_detections_num);
// text output
qsort(selected_detections, selected_detections_num, sizeof(*selected_detections), compare_by_lefts);
for (int i = 0; i < selected_detections_num; ++i) {
const int best_class = selected_detections[i].best_class;
printf("%s: %.0f%%", names[best_class], selected_detections[i].det.prob[best_class] * 100);
if (ext_output)
printf("\t(left: %.0f\ttop: %.0f\tw: %0.f\th: %0.f)\n",
(selected_detections[i].det.bbox.x - selected_detections[i].det.bbox.w / 2)*im.w,
(selected_detections[i].det.bbox.y - selected_detections[i].det.bbox.h / 2)*im.h,
selected_detections[i].det.bbox.w*im.w, selected_detections[i].det.bbox.h*im.h);
else
printf("\n");
for (int j = 0; j < classes; ++j) {
if (selected_detections[i].det.prob[j] > thresh && j != best_class) {
printf("%s: %.0f%%\n", names[j], selected_detections[i].det.prob[j] * 100);
}
}
}
// image output
qsort(selected_detections, selected_detections_num, sizeof(*selected_detections), compare_by_probs);
for (int i = 0; i < selected_detections_num; ++i) {
int width = im.h * .006;
if (width < 1)
width = 1;
@ -262,8 +311,8 @@ void draw_detections_v3(image im, detection *dets, int num, float thresh, char *
}
*/
//printf("%d %s: %.0f%%\n", i, names[class_id], prob*100);
int offset = class_id * 123457 % classes;
//printf("%d %s: %.0f%%\n", i, names[selected_detections[i].best_class], prob*100);
int offset = selected_detections[i].best_class * 123457 % classes;
float red = get_color(2, offset, classes);
float green = get_color(1, offset, classes);
float blue = get_color(0, offset, classes);
@ -274,7 +323,7 @@ void draw_detections_v3(image im, detection *dets, int num, float thresh, char *
rgb[0] = red;
rgb[1] = green;
rgb[2] = blue;
box b = dets[i].bbox;
box b = selected_detections[i].det.bbox;
//printf("%f %f %f %f\n", b.x, b.y, b.w, b.h);
int left = (b.x - b.w / 2.)*im.w;
@ -295,12 +344,20 @@ void draw_detections_v3(image im, detection *dets, int num, float thresh, char *
draw_box_width(im, left, top, right, bot, width, red, green, blue);
if (alphabet) {
char labelstr[4096] = { 0 };
strcat(labelstr, names[selected_detections[i].best_class]);
for (int j = 0; j < classes; ++j) {
if (selected_detections[i].det.prob[j] > thresh && j != selected_detections[i].best_class) {
strcat(labelstr, ", ");
strcat(labelstr, names[j]);
}
}
image label = get_label_v3(alphabet, labelstr, (im.h*.03));
draw_label(im, top + width, left, label, rgb);
free_image(label);
}
if (dets[i].mask) {
image mask = float_to_image(14, 14, 1, dets[i].mask);
if (selected_detections[i].det.mask) {
image mask = float_to_image(14, 14, 1, selected_detections[i].det.mask);
image resized_mask = resize_image(mask, b.w*im.w, b.h*im.h);
image tmask = threshold_image(resized_mask, .5);
embed_image(tmask, im, left, top);
@ -308,8 +365,8 @@ void draw_detections_v3(image im, detection *dets, int num, float thresh, char *
free_image(resized_mask);
free_image(tmask);
}
}
}
free(selected_detections);
}
void draw_detections(image im, int num, float thresh, box *boxes, float **probs, char **names, image **alphabet, int classes)

2
src/image.h
Unescape View File

@ -23,7 +23,7 @@ void draw_bbox(image a, box bbox, int w, float r, float g, float b);
void draw_label(image a, int r, int c, image label, const float *rgb);
void write_label(image a, int r, int c, image *characters, char *string, float *rgb);
void draw_detections(image im, int num, float thresh, box *boxes, float **probs, char **names, image **labels, int classes);
void draw_detections_v3(image im, detection *dets, int num, float thresh, char **names, image **alphabet, int classes);
void draw_detections_v3(image im, detection *dets, int num, float thresh, char **names, image **alphabet, int classes, int ext_output);
image image_distance(image a, image b);
void scale_image(image m, float s);
image crop_image(image im, int dx, int dy, int w, int h);

Write Preview
Loading…
Cancel
Save