From 59e356730fa62d1e307d4fc7a5657f8492633088 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Tue, 22 Sep 2015 22:24:54 +0000
Subject: [PATCH] writing stuff
---
src/region_layer.c | 203 ++++++++++++++++++++++++++------------------------
1 files changed, 104 insertions(+), 99 deletions(-)
diff --git a/src/region_layer.c b/src/region_layer.c
index dcdcfad..39af5ee 100644
--- a/src/region_layer.c
+++ b/src/region_layer.c
@@ -14,7 +14,7 @@
{
region_layer l = {0};
l.type = REGION;
-
+
l.n = n;
l.batch = batch;
l.inputs = inputs;
@@ -22,15 +22,15 @@
l.coords = coords;
l.rescore = rescore;
l.side = side;
- assert(side*side*l.coords*l.n == inputs);
+ assert(side*side*((1 + l.coords)*l.n + l.classes) == inputs);
l.cost = calloc(1, sizeof(float));
- int outputs = l.n*5*side*side;
- l.outputs = outputs;
- l.output = calloc(batch*outputs, sizeof(float));
- l.delta = calloc(batch*inputs, sizeof(float));
- #ifdef GPU
- l.output_gpu = cuda_make_array(l.output, batch*outputs);
- l.delta_gpu = cuda_make_array(l.delta, batch*inputs);
+ l.outputs = l.inputs;
+ l.truths = l.side*l.side*(1+l.coords+l.classes);
+ l.output = calloc(batch*l.outputs, sizeof(float));
+ l.delta = calloc(batch*l.outputs, sizeof(float));
+#ifdef GPU
+ l.output_gpu = cuda_make_array(l.output, batch*l.outputs);
+ l.delta_gpu = cuda_make_array(l.delta, batch*l.outputs);
#endif
fprintf(stderr, "Region Layer\n");
@@ -43,119 +43,123 @@
{
int locations = l.side*l.side;
int i,j;
- for(i = 0; i < l.batch*locations; ++i){
- for(j = 0; j < l.n; ++j){
- int in_index = i*l.n*l.coords + j*l.coords;
- int out_index = i*l.n*5 + j*5;
-
- float prob = state.input[in_index+0];
- float x = state.input[in_index+1];
- float y = state.input[in_index+2];
- float w = state.input[in_index+3];
- float h = state.input[in_index+4];
- /*
- float min_w = state.input[in_index+5];
- float max_w = state.input[in_index+6];
- float min_h = state.input[in_index+7];
- float max_h = state.input[in_index+8];
- */
-
- l.output[out_index+0] = prob;
- l.output[out_index+1] = x;
- l.output[out_index+2] = y;
- l.output[out_index+3] = w;
- l.output[out_index+4] = h;
-
+ memcpy(l.output, state.input, l.outputs*l.batch*sizeof(float));
+ int b;
+ if (l.softmax){
+ for(b = 0; b < l.batch; ++b){
+ int index = b*l.inputs;
+ for (i = 0; i < locations; ++i) {
+ int offset = i*l.classes;
+ softmax_array(l.output + index + offset, l.classes,
+ l.output + index + offset);
+ }
+ int offset = locations*l.classes;
+ activate_array(l.output + index + offset, locations*l.n*(1+l.coords), LOGISTIC);
}
}
+
if(state.train){
float avg_iou = 0;
+ float avg_cat = 0;
+ float avg_allcat = 0;
+ float avg_obj = 0;
+ float avg_anyobj = 0;
int count = 0;
*(l.cost) = 0;
int size = l.inputs * l.batch;
memset(l.delta, 0, size * sizeof(float));
- for (i = 0; i < l.batch*locations; ++i) {
+ for (b = 0; b < l.batch; ++b){
+ int index = b*l.inputs;
+ for (i = 0; i < locations; ++i) {
+ int truth_index = (b*locations + i)*(1+l.coords+l.classes);
+ int is_obj = state.truth[truth_index];
+ for (j = 0; j < l.n; ++j) {
+ int p_index = index + locations*l.classes + i*l.n + j;
+ l.delta[p_index] = l.noobject_scale*(0 - l.output[p_index]);
+ *(l.cost) += l.noobject_scale*pow(l.output[p_index], 2);
+ avg_anyobj += l.output[p_index];
+ }
- for(j = 0; j < l.n; ++j){
- int in_index = i*l.n*l.coords + j*l.coords;
- l.delta[in_index+0] = .1*(0-state.input[in_index+0]);
- }
+ int best_index = -1;
+ float best_iou = 0;
+ float best_rmse = 4;
- int truth_index = i*5;
- int best_index = -1;
- float best_iou = 0;
- float best_rmse = 4;
+ if (!is_obj) continue;
- int bg = !state.truth[truth_index];
- if(bg) continue;
+ int class_index = index + i*l.classes;
+ for(j = 0; j < l.classes; ++j) {
+ l.delta[class_index+j] = l.class_scale * (state.truth[truth_index+1+j] - l.output[class_index+j]);
+ *(l.cost) += l.class_scale * pow(state.truth[truth_index+1+j] - l.output[class_index+j], 2);
+ if(state.truth[truth_index + 1 + j]) avg_cat += l.output[class_index+j];
+ avg_allcat += l.output[class_index+j];
+ }
- box truth = {state.truth[truth_index+1], state.truth[truth_index+2], state.truth[truth_index+3], state.truth[truth_index+4]};
- truth.x /= l.side;
- truth.y /= l.side;
+ box truth = float_to_box(state.truth + truth_index + 1 + l.classes);
+ truth.x /= l.side;
+ truth.y /= l.side;
- for(j = 0; j < l.n; ++j){
- int out_index = i*l.n*5 + j*5;
- box out = {l.output[out_index+1], l.output[out_index+2], l.output[out_index+3], l.output[out_index+4]};
+ for(j = 0; j < l.n; ++j){
+ int box_index = index + locations*(l.classes + l.n) + (i*l.n + j) * l.coords;
+ box out = float_to_box(l.output + box_index);
+ out.x /= l.side;
+ out.y /= l.side;
- //printf("\n%f %f %f %f %f\n", l.output[out_index+0], out.x, out.y, out.w, out.h);
-
- out.x /= l.side;
- out.y /= l.side;
-
- float iou = box_iou(out, truth);
- float rmse = box_rmse(out, truth);
- if(best_iou > 0 || iou > 0){
- if(iou > best_iou){
- best_iou = iou;
- best_index = j;
+ if (l.sqrt){
+ out.w = out.w*out.w;
+ out.h = out.h*out.h;
}
- }else{
- if(rmse < best_rmse){
- best_rmse = rmse;
- best_index = j;
+
+ float iou = box_iou(out, truth);
+ float rmse = box_rmse(out, truth);
+ if(best_iou > 0 || iou > 0){
+ if(iou > best_iou){
+ best_iou = iou;
+ best_index = j;
+ }
+ }else{
+ if(rmse < best_rmse){
+ best_rmse = rmse;
+ best_index = j;
+ }
}
}
+ int p_index = index + locations*l.classes + i*l.n + best_index;
+ *(l.cost) -= l.noobject_scale * pow(l.output[p_index], 2);
+ *(l.cost) += l.object_scale * pow(1-l.output[p_index], 2);
+ avg_obj += l.output[p_index];
+ l.delta[p_index+0] = l.object_scale * (1.-l.output[p_index]);
+
+ if(l.rescore){
+ l.delta[p_index+0] = l.object_scale * (best_iou - l.output[p_index]);
+ }
+
+ int box_index = index + locations*(l.classes + l.n) + (i*l.n + best_index) * l.coords;
+ int tbox_index = truth_index + 1 + l.classes;
+ l.delta[box_index+0] = l.coord_scale*(state.truth[tbox_index + 0] - l.output[box_index + 0]);
+ l.delta[box_index+1] = l.coord_scale*(state.truth[tbox_index + 1] - l.output[box_index + 1]);
+ l.delta[box_index+2] = l.coord_scale*(state.truth[tbox_index + 2] - l.output[box_index + 2]);
+ l.delta[box_index+3] = l.coord_scale*(state.truth[tbox_index + 3] - l.output[box_index + 3]);
+ if(l.sqrt){
+ l.delta[box_index+2] = l.coord_scale*(sqrt(state.truth[tbox_index + 2]) - l.output[box_index + 2]);
+ l.delta[box_index+3] = l.coord_scale*(sqrt(state.truth[tbox_index + 3]) - l.output[box_index + 3]);
+ }
+
+ *(l.cost) += pow(1-best_iou, 2);
+ avg_iou += best_iou;
+ ++count;
}
- printf("%d", best_index);
- //int out_index = i*l.n*5 + best_index*5;
- //box out = {l.output[out_index+1], l.output[out_index+2], l.output[out_index+3], l.output[out_index+4]};
- int in_index = i*l.n*l.coords + best_index*l.coords;
-
- l.delta[in_index+0] = (1-state.input[in_index+0]);
- l.delta[in_index+1] = state.truth[truth_index+1] - state.input[in_index+1];
- l.delta[in_index+2] = state.truth[truth_index+2] - state.input[in_index+2];
- l.delta[in_index+3] = state.truth[truth_index+3] - state.input[in_index+3];
- l.delta[in_index+4] = state.truth[truth_index+4] - state.input[in_index+4];
- /*
- l.delta[in_index+5] = 0 - state.input[in_index+5];
- l.delta[in_index+6] = 1 - state.input[in_index+6];
- l.delta[in_index+7] = 0 - state.input[in_index+7];
- l.delta[in_index+8] = 1 - state.input[in_index+8];
- */
-
- /*
- float x = state.input[in_index+1];
- float y = state.input[in_index+2];
- float w = state.input[in_index+3];
- float h = state.input[in_index+4];
- float min_w = state.input[in_index+5];
- float max_w = state.input[in_index+6];
- float min_h = state.input[in_index+7];
- float max_h = state.input[in_index+8];
- */
-
-
- avg_iou += best_iou;
- ++count;
+ if(l.softmax){
+ gradient_array(l.output + index + locations*l.classes, locations*l.n*(1+l.coords),
+ LOGISTIC, l.delta + index + locations*l.classes);
+ }
}
- printf("\nAvg IOU: %f %d\n", avg_iou/count, count);
+ printf("Region Avg IOU: %f, Pos Cat: %f, All Cat: %f, Pos Obj: %f, Any Obj: %f, count: %d\n", avg_iou/count, avg_cat/count, avg_allcat/(count*l.classes), avg_obj/count, avg_anyobj/(l.batch*locations*l.n), count);
}
}
void backward_region_layer(const region_layer l, network_state state)
{
axpy_cpu(l.batch*l.inputs, 1, l.delta, 1, state.delta, 1);
- //copy_cpu(l.batch*l.inputs, l.delta, 1, state.delta, 1);
}
#ifdef GPU
@@ -165,8 +169,9 @@
float *in_cpu = calloc(l.batch*l.inputs, sizeof(float));
float *truth_cpu = 0;
if(state.truth){
- truth_cpu = calloc(l.batch*l.outputs, sizeof(float));
- cuda_pull_array(state.truth, truth_cpu, l.batch*l.outputs);
+ int num_truth = l.batch*l.side*l.side*(1+l.coords+l.classes);
+ truth_cpu = calloc(num_truth, sizeof(float));
+ cuda_pull_array(state.truth, truth_cpu, num_truth);
}
cuda_pull_array(state.input, in_cpu, l.batch*l.inputs);
network_state cpu_state;
--
Gitblit v1.10.0