From 8561e49b5a2876e9a522b2dedfa99f19d5738154 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Mon, 13 Jul 2015 22:04:21 +0000
Subject: [PATCH] add avgpool layer
---
src/detection_layer.c | 213 +++++++++++++++++++++++++++++++++++++---------------
1 files changed, 150 insertions(+), 63 deletions(-)
diff --git a/src/detection_layer.c b/src/detection_layer.c
index bbc2e4f..9ef89d9 100644
--- a/src/detection_layer.c
+++ b/src/detection_layer.c
@@ -2,119 +2,206 @@
#include "activations.h"
#include "softmax_layer.h"
#include "blas.h"
+#include "box.h"
#include "cuda.h"
+#include "utils.h"
#include <stdio.h>
+#include <string.h>
#include <stdlib.h>
-int get_detection_layer_locations(detection_layer layer)
+int get_detection_layer_locations(detection_layer l)
{
- return layer.inputs / (layer.classes+layer.coords+layer.rescore);
+ return l.inputs / (l.classes+l.coords+l.joint+(l.background || l.objectness));
}
-int get_detection_layer_output_size(detection_layer layer)
+int get_detection_layer_output_size(detection_layer l)
{
- return get_detection_layer_locations(layer)*(layer.classes+layer.coords);
+ return get_detection_layer_locations(l)*((l.background || l.objectness) + l.classes + l.coords);
}
-detection_layer *make_detection_layer(int batch, int inputs, int classes, int coords, int rescore)
+detection_layer make_detection_layer(int batch, int inputs, int classes, int coords, int joint, int rescore, int background, int objectness)
{
- detection_layer *layer = calloc(1, sizeof(detection_layer));
-
- layer->batch = batch;
- layer->inputs = inputs;
- layer->classes = classes;
- layer->coords = coords;
- layer->rescore = rescore;
- int outputs = get_detection_layer_output_size(*layer);
- layer->output = calloc(batch*outputs, sizeof(float));
- layer->delta = calloc(batch*outputs, sizeof(float));
+ detection_layer l = {0};
+ l.type = DETECTION;
+
+ l.batch = batch;
+ l.inputs = inputs;
+ l.classes = classes;
+ l.coords = coords;
+ l.rescore = rescore;
+ l.objectness = objectness;
+ l.background = background;
+ l.joint = joint;
+ l.cost = calloc(1, sizeof(float));
+ l.does_cost=1;
+ int outputs = get_detection_layer_output_size(l);
+ l.outputs = outputs;
+ l.output = calloc(batch*outputs, sizeof(float));
+ l.delta = calloc(batch*outputs, sizeof(float));
#ifdef GPU
- layer->output_gpu = cuda_make_array(0, batch*outputs);
- layer->delta_gpu = cuda_make_array(0, batch*outputs);
+ l.output_gpu = cuda_make_array(0, batch*outputs);
+ l.delta_gpu = cuda_make_array(0, batch*outputs);
#endif
fprintf(stderr, "Detection Layer\n");
srand(0);
- return layer;
+ return l;
}
-void forward_detection_layer(const detection_layer layer, float *in, float *truth)
+void forward_detection_layer(const detection_layer l, network_state state)
{
int in_i = 0;
int out_i = 0;
- int locations = get_detection_layer_locations(layer);
+ int locations = get_detection_layer_locations(l);
int i,j;
- for(i = 0; i < layer.batch*locations; ++i){
- int mask = (!truth || !truth[out_i + layer.classes - 1]);
+ for(i = 0; i < l.batch*locations; ++i){
+ int mask = (!state.truth || state.truth[out_i + (l.background || l.objectness) + l.classes + 2]);
float scale = 1;
- if(layer.rescore) scale = in[in_i++];
- for(j = 0; j < layer.classes; ++j){
- layer.output[out_i++] = scale*in[in_i++];
+ if(l.joint) scale = state.input[in_i++];
+ else if(l.objectness){
+ l.output[out_i++] = 1-state.input[in_i++];
+ scale = mask;
}
- softmax_array(layer.output + out_i - layer.classes, layer.classes, layer.output + out_i - layer.classes);
- activate_array(layer.output+out_i, layer.coords, SIGMOID);
- for(j = 0; j < layer.coords; ++j){
- layer.output[out_i++] = mask*in[in_i++];
+ else if(l.background) l.output[out_i++] = scale*state.input[in_i++];
+
+ for(j = 0; j < l.classes; ++j){
+ l.output[out_i++] = scale*state.input[in_i++];
}
- //printf("%d\n", mask);
- //for(j = 0; j < layer.classes+layer.coords; ++j) printf("%f ", layer.output[i*(layer.classes+layer.coords)+j]);
- //printf ("\n");
+ if(l.objectness){
+
+ }else if(l.background){
+ softmax_array(l.output + out_i - l.classes-l.background, l.classes+l.background, l.output + out_i - l.classes-l.background);
+ activate_array(state.input+in_i, l.coords, LOGISTIC);
+ }
+ for(j = 0; j < l.coords; ++j){
+ l.output[out_i++] = mask*state.input[in_i++];
+ }
+ }
+ float avg_iou = 0;
+ int count = 0;
+ if(l.does_cost && state.train){
+ *(l.cost) = 0;
+ int size = get_detection_layer_output_size(l) * l.batch;
+ memset(l.delta, 0, size * sizeof(float));
+ for (i = 0; i < l.batch*locations; ++i) {
+ int classes = l.objectness+l.classes;
+ int offset = i*(classes+l.coords);
+ for (j = offset; j < offset+classes; ++j) {
+ *(l.cost) += pow(state.truth[j] - l.output[j], 2);
+ l.delta[j] = state.truth[j] - l.output[j];
+ }
+
+ box truth;
+ truth.x = state.truth[j+0]/7;
+ truth.y = state.truth[j+1]/7;
+ truth.w = pow(state.truth[j+2], 2);
+ truth.h = pow(state.truth[j+3], 2);
+ box out;
+ out.x = l.output[j+0]/7;
+ out.y = l.output[j+1]/7;
+ out.w = pow(l.output[j+2], 2);
+ out.h = pow(l.output[j+3], 2);
+
+ if(!(truth.w*truth.h)) continue;
+ float iou = box_iou(out, truth);
+ avg_iou += iou;
+ ++count;
+ dbox delta = diou(out, truth);
+
+ l.delta[j+0] = 10 * delta.dx/7;
+ l.delta[j+1] = 10 * delta.dy/7;
+ l.delta[j+2] = 10 * delta.dw * 2 * sqrt(out.w);
+ l.delta[j+3] = 10 * delta.dh * 2 * sqrt(out.h);
+
+
+ *(l.cost) += pow((1-iou), 2);
+ l.delta[j+0] = 4 * (state.truth[j+0] - l.output[j+0]);
+ l.delta[j+1] = 4 * (state.truth[j+1] - l.output[j+1]);
+ l.delta[j+2] = 4 * (state.truth[j+2] - l.output[j+2]);
+ l.delta[j+3] = 4 * (state.truth[j+3] - l.output[j+3]);
+ if(l.rescore){
+ for (j = offset; j < offset+classes; ++j) {
+ if(state.truth[j]) state.truth[j] = iou;
+ l.delta[j] = state.truth[j] - l.output[j];
+ }
+ }
+ }
+ printf("Avg IOU: %f\n", avg_iou/count);
}
}
-void backward_detection_layer(const detection_layer layer, float *in, float *delta)
+void backward_detection_layer(const detection_layer l, network_state state)
{
- int locations = get_detection_layer_locations(layer);
+ int locations = get_detection_layer_locations(l);
int i,j;
int in_i = 0;
int out_i = 0;
- for(i = 0; i < layer.batch*locations; ++i){
+ for(i = 0; i < l.batch*locations; ++i){
float scale = 1;
float latent_delta = 0;
- if(layer.rescore) scale = in[in_i++];
- for(j = 0; j < layer.classes; ++j){
- latent_delta += in[in_i]*layer.delta[out_i];
- delta[in_i++] = scale*layer.delta[out_i++];
+ if(l.joint) scale = state.input[in_i++];
+ else if (l.objectness) state.delta[in_i++] = -l.delta[out_i++];
+ else if (l.background) state.delta[in_i++] = scale*l.delta[out_i++];
+ for(j = 0; j < l.classes; ++j){
+ latent_delta += state.input[in_i]*l.delta[out_i];
+ state.delta[in_i++] = scale*l.delta[out_i++];
}
-
- for(j = 0; j < layer.coords; ++j){
- delta[in_i++] = layer.delta[out_i++];
+
+ if (l.objectness) {
+
+ }else if (l.background) gradient_array(l.output + out_i, l.coords, LOGISTIC, l.delta + out_i);
+ for(j = 0; j < l.coords; ++j){
+ state.delta[in_i++] = l.delta[out_i++];
}
- gradient_array(in + in_i - layer.coords, layer.coords, SIGMOID, layer.delta + out_i - layer.coords);
- if(layer.rescore) delta[in_i-layer.coords-layer.classes-layer.rescore] = latent_delta;
+ if(l.joint) state.delta[in_i-l.coords-l.classes-l.joint] = latent_delta;
}
}
#ifdef GPU
-void forward_detection_layer_gpu(const detection_layer layer, float *in, float *truth)
+void forward_detection_layer_gpu(const detection_layer l, network_state state)
{
- int outputs = get_detection_layer_output_size(layer);
- float *in_cpu = calloc(layer.batch*layer.inputs, sizeof(float));
+ int outputs = get_detection_layer_output_size(l);
+ float *in_cpu = calloc(l.batch*l.inputs, sizeof(float));
float *truth_cpu = 0;
- if(truth){
- truth_cpu = calloc(layer.batch*outputs, sizeof(float));
- cuda_pull_array(truth, truth_cpu, layer.batch*outputs);
+ if(state.truth){
+ truth_cpu = calloc(l.batch*outputs, sizeof(float));
+ cuda_pull_array(state.truth, truth_cpu, l.batch*outputs);
}
- cuda_pull_array(in, in_cpu, layer.batch*layer.inputs);
- forward_detection_layer(layer, in_cpu, truth_cpu);
- cuda_push_array(layer.output_gpu, layer.output, layer.batch*outputs);
- free(in_cpu);
- if(truth_cpu) free(truth_cpu);
+ cuda_pull_array(state.input, in_cpu, l.batch*l.inputs);
+ network_state cpu_state;
+ cpu_state.train = state.train;
+ cpu_state.truth = truth_cpu;
+ cpu_state.input = in_cpu;
+ forward_detection_layer(l, cpu_state);
+ cuda_push_array(l.output_gpu, l.output, l.batch*outputs);
+ cuda_push_array(l.delta_gpu, l.delta, l.batch*outputs);
+ free(cpu_state.input);
+ if(cpu_state.truth) free(cpu_state.truth);
}
-void backward_detection_layer_gpu(detection_layer layer, float *in, float *delta)
+void backward_detection_layer_gpu(detection_layer l, network_state state)
{
- int outputs = get_detection_layer_output_size(layer);
+ int outputs = get_detection_layer_output_size(l);
- float *in_cpu = calloc(layer.batch*layer.inputs, sizeof(float));
- float *delta_cpu = calloc(layer.batch*layer.inputs, sizeof(float));
+ float *in_cpu = calloc(l.batch*l.inputs, sizeof(float));
+ float *delta_cpu = calloc(l.batch*l.inputs, sizeof(float));
+ float *truth_cpu = 0;
+ if(state.truth){
+ truth_cpu = calloc(l.batch*outputs, sizeof(float));
+ cuda_pull_array(state.truth, truth_cpu, l.batch*outputs);
+ }
+ network_state cpu_state;
+ cpu_state.train = state.train;
+ cpu_state.input = in_cpu;
+ cpu_state.truth = truth_cpu;
+ cpu_state.delta = delta_cpu;
- cuda_pull_array(in, in_cpu, layer.batch*layer.inputs);
- cuda_pull_array(layer.delta_gpu, layer.delta, layer.batch*outputs);
- backward_detection_layer(layer, in_cpu, delta_cpu);
- cuda_push_array(delta, delta_cpu, layer.batch*layer.inputs);
+ cuda_pull_array(state.input, in_cpu, l.batch*l.inputs);
+ cuda_pull_array(l.delta_gpu, l.delta, l.batch*outputs);
+ backward_detection_layer(l, cpu_state);
+ cuda_push_array(state.delta, delta_cpu, l.batch*l.inputs);
free(in_cpu);
free(delta_cpu);
--
Gitblit v1.10.0