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

counters_per_class affects the bbox too

Browse Source
pull/1557/head
AlexeyAB 5 years ago
parent f64746107d
commit b8fe630119
2 changed files with 24 additions and 14 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 19
      src/gaussian_yolo_layer.c
  2. 19
      src/yolo_layer.c

19
src/gaussian_yolo_layer.c
Unescape View File

@ -456,7 +456,8 @@ void forward_gaussian_yolo_layer(const layer l, network_state state)
int class_index = entry_gaussian_index(l, b, n*l.w*l.h + j*l.w + i, 9); int class_index = entry_gaussian_index(l, b, n*l.w*l.h + j*l.w + i, 9);
delta_gaussian_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, 0, l.label_smooth_eps, l.classes_multipliers); delta_gaussian_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, 0, l.label_smooth_eps, l.classes_multipliers);
box truth = float_to_box_stride(state.truth + best_t*(4 + 1) + b*l.truths, 1); box truth = float_to_box_stride(state.truth + best_t*(4 + 1) + b*l.truths, 1);
delta_gaussian_yolo_box(truth, l.output, l.biases, l.mask[n], box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2-truth.w*truth.h), l.w*l.h, l.iou_normalizer, l.iou_loss, l.uc_normalizer, 1, l.yolo_point); const float class_multiplier = (l.classes_multipliers) ? l.classes_multipliers[class_id] : 1.0f;
delta_gaussian_yolo_box(truth, l.output, l.biases, l.mask[n], box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2-truth.w*truth.h), l.w*l.h, l.iou_normalizer * class_multiplier, l.iou_loss, l.uc_normalizer, 1, l.yolo_point);
} }
} }
} }
@ -496,15 +497,17 @@ void forward_gaussian_yolo_layer(const layer l, network_state state)
int mask_n = int_index(l.mask, best_n, l.n); int mask_n = int_index(l.mask, best_n, l.n);
if(mask_n >= 0){ if(mask_n >= 0){
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int box_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0); int box_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0);
float iou = delta_gaussian_yolo_box(truth, l.output, l.biases, best_n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2-truth.w*truth.h), l.w*l.h, l.iou_normalizer, l.iou_loss, l.uc_normalizer, 1, l.yolo_point); const float class_multiplier = (l.classes_multipliers) ? l.classes_multipliers[class_id] : 1.0f;
float iou = delta_gaussian_yolo_box(truth, l.output, l.biases, best_n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2-truth.w*truth.h), l.w*l.h, l.iou_normalizer * class_multiplier, l.iou_loss, l.uc_normalizer, 1, l.yolo_point);
int obj_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 8); int obj_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 8);
avg_obj += l.output[obj_index]; avg_obj += l.output[obj_index];
l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]); l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]);
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int class_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 9); int class_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 9);
delta_gaussian_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.label_smooth_eps, l.classes_multipliers); delta_gaussian_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.label_smooth_eps, l.classes_multipliers);
@ -527,15 +530,17 @@ void forward_gaussian_yolo_layer(const layer l, network_state state)
// iou, n // iou, n
if (iou > l.iou_thresh) { if (iou > l.iou_thresh) {
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int box_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0); int box_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0);
float iou = delta_gaussian_yolo_box(truth, l.output, l.biases, n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer, l.iou_loss, l.uc_normalizer, 1, l.yolo_point); const float class_multiplier = (l.classes_multipliers) ? l.classes_multipliers[class_id] : 1.0f;
float iou = delta_gaussian_yolo_box(truth, l.output, l.biases, n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer * class_multiplier, l.iou_loss, l.uc_normalizer, 1, l.yolo_point);
int obj_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 8); int obj_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 8);
avg_obj += l.output[obj_index]; avg_obj += l.output[obj_index];
l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]); l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]);
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int class_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 9); int class_index = entry_gaussian_index(l, b, mask_n*l.w*l.h + j*l.w + i, 9);
delta_gaussian_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.label_smooth_eps, l.classes_multipliers); delta_gaussian_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.label_smooth_eps, l.classes_multipliers);

19
src/yolo_layer.c
Unescape View File

@ -363,7 +363,8 @@ void forward_yolo_layer(const layer l, network_state state)
int class_index = entry_index(l, b, n*l.w*l.h + j*l.w + i, 4 + 1); int class_index = entry_index(l, b, n*l.w*l.h + j*l.w + i, 4 + 1);
delta_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, 0, l.focal_loss, l.label_smooth_eps, l.classes_multipliers); delta_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, 0, l.focal_loss, l.label_smooth_eps, l.classes_multipliers);
box truth = float_to_box_stride(state.truth + best_t*(4 + 1) + b*l.truths, 1); box truth = float_to_box_stride(state.truth + best_t*(4 + 1) + b*l.truths, 1);
delta_yolo_box(truth, l.output, l.biases, l.mask[n], box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer, l.iou_loss, 1); const float class_multiplier = (l.classes_multipliers) ? l.classes_multipliers[class_id] : 1.0f;
delta_yolo_box(truth, l.output, l.biases, l.mask[n], box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer * class_multiplier, l.iou_loss, 1);
} }
} }
} }
@ -400,8 +401,12 @@ void forward_yolo_layer(const layer l, network_state state)
int mask_n = int_index(l.mask, best_n, l.n); int mask_n = int_index(l.mask, best_n, l.n);
if (mask_n >= 0) { if (mask_n >= 0) {
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int box_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0); int box_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0);
ious all_ious = delta_yolo_box(truth, l.output, l.biases, best_n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer, l.iou_loss, 1); const float class_multiplier = (l.classes_multipliers) ? l.classes_multipliers[class_id] : 1.0f;
ious all_ious = delta_yolo_box(truth, l.output, l.biases, best_n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer * class_multiplier, l.iou_loss, 1);
// range is 0 <= 1 // range is 0 <= 1
tot_iou += all_ious.iou; tot_iou += all_ious.iou;
@ -420,8 +425,6 @@ void forward_yolo_layer(const layer l, network_state state)
avg_obj += l.output[obj_index]; avg_obj += l.output[obj_index];
l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]); l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]);
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int class_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 4 + 1); int class_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 4 + 1);
delta_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.focal_loss, l.label_smooth_eps, l.classes_multipliers); delta_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.focal_loss, l.label_smooth_eps, l.classes_multipliers);
@ -442,8 +445,12 @@ void forward_yolo_layer(const layer l, network_state state)
// iou, n // iou, n
if (iou > l.iou_thresh) { if (iou > l.iou_thresh) {
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int box_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0); int box_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 0);
ious all_ious = delta_yolo_box(truth, l.output, l.biases, n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer, l.iou_loss, 1); const float class_multiplier = (l.classes_multipliers) ? l.classes_multipliers[class_id] : 1.0f;
ious all_ious = delta_yolo_box(truth, l.output, l.biases, n, box_index, i, j, l.w, l.h, state.net.w, state.net.h, l.delta, (2 - truth.w*truth.h), l.w*l.h, l.iou_normalizer * class_multiplier, l.iou_loss, 1);
// range is 0 <= 1 // range is 0 <= 1
tot_iou += all_ious.iou; tot_iou += all_ious.iou;
@ -462,8 +469,6 @@ void forward_yolo_layer(const layer l, network_state state)
avg_obj += l.output[obj_index]; avg_obj += l.output[obj_index];
l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]); l.delta[obj_index] = l.cls_normalizer * (1 - l.output[obj_index]);
int class_id = state.truth[t*(4 + 1) + b*l.truths + 4];
if (l.map) class_id = l.map[class_id];
int class_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 4 + 1); int class_index = entry_index(l, b, mask_n*l.w*l.h + j*l.w + i, 4 + 1);
delta_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.focal_loss, l.label_smooth_eps, l.classes_multipliers); delta_yolo_class(l.output, l.delta, class_index, class_id, l.classes, l.w*l.h, &avg_cat, l.focal_loss, l.label_smooth_eps, l.classes_multipliers);

Write Preview
Loading…
Cancel
Save