From 252e3b1916cfaca0783c9e90efaa55eb07b1a8cd Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Sun, 06 Nov 2016 00:27:31 +0000
Subject: [PATCH] :charizard: :charizard: :charizard:
---
src/region_layer.c | 238 ++++++++++++++++++++++++++++++++++-------------------------
1 files changed, 138 insertions(+), 100 deletions(-)
diff --git a/src/region_layer.c b/src/region_layer.c
index 2185ea5..2702636 100644
--- a/src/region_layer.c
+++ b/src/region_layer.c
@@ -1,6 +1,5 @@
#include "region_layer.h"
#include "activations.h"
-#include "softmax_layer.h"
#include "blas.h"
#include "box.h"
#include "cuda.h"
@@ -22,12 +21,23 @@
l.classes = classes;
l.coords = coords;
l.cost = calloc(1, sizeof(float));
+ l.biases = calloc(n*2, sizeof(float));
+ l.bias_updates = calloc(n*2, sizeof(float));
l.outputs = h*w*n*(classes + coords + 1);
l.inputs = l.outputs;
l.truths = 30*(5);
l.delta = calloc(batch*l.outputs, sizeof(float));
l.output = calloc(batch*l.outputs, sizeof(float));
+ int i;
+ for(i = 0; i < n*2; ++i){
+ l.biases[i] = .5;
+ }
+
+ l.forward = forward_region_layer;
+ l.backward = backward_region_layer;
#ifdef GPU
+ l.forward_gpu = forward_region_layer_gpu;
+ l.backward_gpu = backward_region_layer_gpu;
l.output_gpu = cuda_make_array(l.output, batch*l.outputs);
l.delta_gpu = cuda_make_array(l.delta, batch*l.outputs);
#endif
@@ -38,62 +48,44 @@
return l;
}
-box get_region_box2(float *x, int index, int i, int j, int w, int h)
-{
- float aspect = exp(x[index+0]);
- float scale = logistic_activate(x[index+1]);
- float move_x = x[index+2];
- float move_y = x[index+3];
+#define LOG 1
- box b;
- b.w = sqrt(scale * aspect);
- b.h = b.w * 1./aspect;
- b.x = move_x * b.w + (i + .5)/w;
- b.y = move_y * b.h + (j + .5)/h;
- return b;
-}
-
-float delta_region_box2(box truth, float *output, int index, int i, int j, int w, int h, float *delta)
-{
- box pred = get_region_box2(output, index, i, j, w, h);
- float iou = box_iou(pred, truth);
- float true_aspect = truth.w/truth.h;
- float true_scale = truth.w*truth.h;
-
- float true_dx = (truth.x - (i+.5)/w) / truth.w;
- float true_dy = (truth.y - (j+.5)/h) / truth.h;
- delta[index + 0] = (true_aspect - exp(output[index + 0])) * exp(output[index + 0]);
- delta[index + 1] = (true_scale - logistic_activate(output[index + 1])) * logistic_gradient(logistic_activate(output[index + 1]));
- delta[index + 2] = true_dx - output[index + 2];
- delta[index + 3] = true_dy - output[index + 3];
- return iou;
-}
-
-box get_region_box(float *x, int index, int i, int j, int w, int h, int adjust, int logistic)
+box get_region_box(float *x, float *biases, int n, int index, int i, int j, int w, int h)
{
box b;
- b.x = (x[index + 0] + i + .5)/w;
- b.y = (x[index + 1] + j + .5)/h;
- b.w = x[index + 2];
- b.h = x[index + 3];
- if(logistic){
- b.w = logistic_activate(x[index + 2]);
- b.h = logistic_activate(x[index + 3]);
+ b.x = (i + .5)/w + x[index + 0] * biases[2*n];
+ b.y = (j + .5)/h + x[index + 1] * biases[2*n + 1];
+ if(LOG){
+ b.x = (i + logistic_activate(x[index + 0])) / w;
+ b.y = (j + logistic_activate(x[index + 1])) / h;
}
- //if(adjust && b.w < .01) b.w = .01;
- //if(adjust && b.h < .01) b.h = .01;
+ b.w = exp(x[index + 2]) * biases[2*n];
+ b.h = exp(x[index + 3]) * biases[2*n+1];
return b;
}
-float delta_region_box(box truth, float *output, int index, int i, int j, int w, int h, float *delta, int logistic, float scale)
+float delta_region_box(box truth, float *x, float *biases, int n, int index, int i, int j, int w, int h, float *delta, float scale)
{
- box pred = get_region_box(output, index, i, j, w, h, 0, logistic);
+ box pred = get_region_box(x, biases, n, index, i, j, w, h);
float iou = box_iou(pred, truth);
- delta[index + 0] = scale * (truth.x - pred.x);
- delta[index + 1] = scale * (truth.y - pred.y);
- delta[index + 2] = scale * ((truth.w - pred.w)*(logistic ? logistic_gradient(pred.w) : 1));
- delta[index + 3] = scale * ((truth.h - pred.h)*(logistic ? logistic_gradient(pred.h) : 1));
+ float tx = (truth.x - (i + .5)/w) / biases[2*n];
+ float ty = (truth.y - (j + .5)/h) / biases[2*n + 1];
+ if(LOG){
+ tx = (truth.x*w - i);
+ ty = (truth.y*h - j);
+ }
+ float tw = log(truth.w / biases[2*n]);
+ float th = log(truth.h / biases[2*n + 1]);
+
+ delta[index + 0] = scale * (tx - x[index + 0]);
+ delta[index + 1] = scale * (ty - x[index + 1]);
+ if(LOG){
+ delta[index + 0] = scale * (tx - logistic_activate(x[index + 0])) * logistic_gradient(logistic_activate(x[index + 0]));
+ delta[index + 1] = scale * (ty - logistic_activate(x[index + 1])) * logistic_gradient(logistic_activate(x[index + 1]));
+ }
+ delta[index + 2] = scale * (tw - x[index + 2]);
+ delta[index + 3] = scale * (th - x[index + 3]);
return iou;
}
@@ -107,8 +99,7 @@
return (x != x);
}
-#define LOG 1
-
+void softmax_tree(float *input, int batch, int inputs, float temp, tree *hierarchy, float *output);
void forward_region_layer(const region_layer l, network_state state)
{
int i,j,b,t,n;
@@ -119,14 +110,17 @@
for(i = 0; i < l.h*l.w*l.n; ++i){
int index = size*i + b*l.outputs;
l.output[index + 4] = logistic_activate(l.output[index + 4]);
- if(l.softmax){
- softmax_array(l.output + index + 5, l.classes, 1, l.output + index + 5);
+ if(l.softmax_tree){
+ softmax_tree(l.output + index + 5, 1, 0, 1, l.softmax_tree, l.output + index + 5);
+ } else if(l.softmax){
+ softmax(l.output + index + 5, l.classes, 1, l.output + index + 5);
}
}
}
if(!state.train) return;
memset(l.delta, 0, l.outputs * l.batch * sizeof(float));
float avg_iou = 0;
+ float recall = 0;
float avg_cat = 0;
float avg_obj = 0;
float avg_anyobj = 0;
@@ -137,7 +131,7 @@
for (i = 0; i < l.w; ++i) {
for (n = 0; n < l.n; ++n) {
int index = size*(j*l.w*l.n + i*l.n + n) + b*l.outputs;
- box pred = get_region_box(l.output, index, i, j, l.w, l.h, 1, LOG);
+ box pred = get_region_box(l.output, l.biases, n, index, i, j, l.w, l.h);
float best_iou = 0;
for(t = 0; t < 30; ++t){
box truth = float_to_box(state.truth + t*5 + b*l.truths);
@@ -147,24 +141,26 @@
}
avg_anyobj += l.output[index + 4];
l.delta[index + 4] = l.noobject_scale * ((0 - l.output[index + 4]) * logistic_gradient(l.output[index + 4]));
- if(best_iou > .5) l.delta[index + 4] = 0;
+ if(best_iou > l.thresh) l.delta[index + 4] = 0;
- /*
- if(*(state.net.seen) < 6400){
+ if(*(state.net.seen) < 12800){
box truth = {0};
truth.x = (i + .5)/l.w;
truth.y = (j + .5)/l.h;
- truth.w = .5;
- truth.h = .5;
- delta_region_box(truth, l.output, index, i, j, l.w, l.h, l.delta, LOG, 1);
+ truth.w = l.biases[2*n];
+ truth.h = l.biases[2*n+1];
+ delta_region_box(truth, l.output, l.biases, n, index, i, j, l.w, l.h, l.delta, .01);
+ //l.delta[index + 0] = .1 * (0 - l.output[index + 0]);
+ //l.delta[index + 1] = .1 * (0 - l.output[index + 1]);
+ //l.delta[index + 2] = .1 * (0 - l.output[index + 2]);
+ //l.delta[index + 3] = .1 * (0 - l.output[index + 3]);
}
- */
}
}
}
for(t = 0; t < 30; ++t){
box truth = float_to_box(state.truth + t*5 + b*l.truths);
- int class = state.truth[t*5 + b*l.truths + 4];
+
if(!truth.x) break;
float best_iou = 0;
int best_index = 0;
@@ -175,11 +171,15 @@
box truth_shift = truth;
truth_shift.x = 0;
truth_shift.y = 0;
- printf("index %d %d\n",i, j);
+ //printf("index %d %d\n",i, j);
for(n = 0; n < l.n; ++n){
int index = size*(j*l.w*l.n + i*l.n + n) + b*l.outputs;
- box pred = get_region_box(l.output, index, i, j, l.w, l.h, 1, LOG);
- printf("pred: (%f, %f) %f x %f\n", pred.x, pred.y, pred.w, pred.h);
+ box pred = get_region_box(l.output, l.biases, n, index, i, j, l.w, l.h);
+ if(l.bias_match){
+ pred.w = l.biases[2*n];
+ pred.h = l.biases[2*n+1];
+ }
+ //printf("pred: (%f, %f) %f x %f\n", pred.x, pred.y, pred.w, pred.h);
pred.x = 0;
pred.y = 0;
float iou = box_iou(pred, truth_shift);
@@ -189,9 +189,10 @@
best_n = n;
}
}
- printf("%d %f (%f, %f) %f x %f\n", best_n, best_iou, truth.x, truth.y, truth.w, truth.h);
+ //printf("%d %f (%f, %f) %f x %f\n", best_n, best_iou, truth.x, truth.y, truth.w, truth.h);
- float iou = delta_region_box(truth, l.output, best_index, i, j, l.w, l.h, l.delta, LOG, l.coord_scale);
+ float iou = delta_region_box(truth, l.output, l.biases, best_n, best_index, i, j, l.w, l.h, l.delta, l.coord_scale);
+ if(iou > .5) recall += 1;
avg_iou += iou;
//l.delta[best_index + 4] = iou - l.output[best_index + 4];
@@ -200,48 +201,39 @@
if (l.rescore) {
l.delta[best_index + 4] = l.object_scale * (iou - l.output[best_index + 4]) * logistic_gradient(l.output[best_index + 4]);
}
- //printf("%f\n", l.delta[best_index+1]);
- /*
- if(isnan(l.delta[best_index+1])){
- printf("%f\n", true_scale);
- printf("%f\n", l.output[best_index + 1]);
- printf("%f\n", truth.w);
- printf("%f\n", truth.h);
- error("bad");
- }
- */
- for(n = 0; n < l.classes; ++n){
- l.delta[best_index + 5 + n] = l.class_scale * (((n == class)?1 : 0) - l.output[best_index + 5 + n]);
- if(n == class) avg_cat += l.output[best_index + 5 + n];
- }
- /*
- if(0){
- printf("truth: %f %f %f %f\n", truth.x, truth.y, truth.w, truth.h);
- printf("pred: %f %f %f %f\n\n", pred.x, pred.y, pred.w, pred.h);
- float aspect = exp(true_aspect);
- float scale = logistic_activate(true_scale);
- float move_x = true_dx;
- float move_y = true_dy;
- box b;
- b.w = sqrt(scale * aspect);
- b.h = b.w * 1./aspect;
- b.x = move_x * b.w + (i + .5)/l.w;
- b.y = move_y * b.h + (j + .5)/l.h;
- printf("%f %f\n", b.x, truth.x);
- printf("%f %f\n", b.y, truth.y);
- printf("%f %f\n", b.w, truth.w);
- printf("%f %f\n", b.h, truth.h);
- //printf("%f\n", box_iou(b, truth));
+
+ int class = state.truth[t*5 + b*l.truths + 4];
+ if (l.map) class = l.map[class];
+ if(l.softmax_tree){
+ float pred = 1;
+ while(class >= 0){
+ pred *= l.output[best_index + 5 + class];
+ int g = l.softmax_tree->group[class];
+ int i;
+ int offset = l.softmax_tree->group_offset[g];
+ for(i = 0; i < l.softmax_tree->group_size[g]; ++i){
+ int index = best_index + 5 + offset + i;
+ l.delta[index] = l.class_scale * (0 - l.output[index]);
+ }
+ l.delta[best_index + 5 + class] = l.class_scale * (1 - l.output[best_index + 5 + class]);
+
+ class = l.softmax_tree->parent[class];
+ }
+ avg_cat += pred;
+ } else {
+ for(n = 0; n < l.classes; ++n){
+ l.delta[best_index + 5 + n] = l.class_scale * (((n == class)?1 : 0) - l.output[best_index + 5 + n]);
+ if(n == class) avg_cat += l.output[best_index + 5 + n];
+ }
}
- */
++count;
}
}
- printf("\n");
+ //printf("\n");
reorg(l.delta, l.w*l.h, size*l.n, l.batch, 0);
*(l.cost) = pow(mag_array(l.delta, l.outputs * l.batch), 2);
- printf("Region Avg IOU: %f, Class: %f, Obj: %f, No Obj: %f, count: %d\n", avg_iou/count, avg_cat/count, avg_obj/count, avg_anyobj/(l.w*l.h*l.n*l.batch), count);
+ printf("Region Avg IOU: %f, Class: %f, Obj: %f, No Obj: %f, Avg Recall: %f, count: %d\n", avg_iou/count, avg_cat/count, avg_obj/count, avg_anyobj/(l.w*l.h*l.n*l.batch), recall/count, count);
}
void backward_region_layer(const region_layer l, network_state state)
@@ -249,6 +241,52 @@
axpy_cpu(l.batch*l.inputs, 1, l.delta, 1, state.delta, 1);
}
+void get_region_boxes(layer l, int w, int h, float thresh, float **probs, box *boxes, int only_objectness)
+{
+ int i,j,n;
+ float *predictions = l.output;
+ //int per_cell = 5*num+classes;
+ for (i = 0; i < l.w*l.h; ++i){
+ int row = i / l.w;
+ int col = i % l.w;
+ for(n = 0; n < l.n; ++n){
+ int index = i*l.n + n;
+ int p_index = index * (l.classes + 5) + 4;
+ float scale = predictions[p_index];
+ int box_index = index * (l.classes + 5);
+ boxes[index] = get_region_box(predictions, l.biases, n, box_index, col, row, l.w, l.h);
+ boxes[index].x *= w;
+ boxes[index].y *= h;
+ boxes[index].w *= w;
+ boxes[index].h *= h;
+
+ int class_index = index * (l.classes + 5) + 5;
+ if(l.softmax_tree){
+
+ hierarchy_predictions(predictions + class_index, l.classes, l.softmax_tree, 0);
+ int found = 0;
+ for(j = l.classes - 1; j >= 0; --j){
+ if(!found && predictions[class_index + j] > .5){
+ found = 1;
+ } else {
+ predictions[class_index + j] = 0;
+ }
+ float prob = predictions[class_index+j];
+ probs[index][j] = (scale > thresh) ? prob : 0;
+ }
+ }else{
+ for(j = 0; j < l.classes; ++j){
+ float prob = scale*predictions[class_index+j];
+ probs[index][j] = (prob > thresh) ? prob : 0;
+ }
+ }
+ if(only_objectness){
+ probs[index][0] = scale;
+ }
+ }
+ }
+}
+
#ifdef GPU
void forward_region_layer_gpu(const region_layer l, network_state state)
--
Gitblit v1.10.0