From 033e934ce82826c73d851098baf7ce4b1a27c89a Mon Sep 17 00:00:00 2001
From: AlexeyAB <alexeyab84@gmail.com>
Date: Wed, 21 Feb 2018 16:14:01 +0000
Subject: [PATCH] If there is excessive GPU-RAM consumption by CUDNN then then do not use Workspace
---
src/detection_layer.c | 332 +++++++++++++++++++++++++++++++++++++++++--------------
1 files changed, 248 insertions(+), 84 deletions(-)
diff --git a/src/detection_layer.c b/src/detection_layer.c
index bbc2e4f..fd5a419 100644
--- a/src/detection_layer.c
+++ b/src/detection_layer.c
@@ -2,122 +2,286 @@
#include "activations.h"
#include "softmax_layer.h"
#include "blas.h"
+#include "box.h"
#include "cuda.h"
+#include "utils.h"
#include <stdio.h>
+#include <assert.h>
+#include <string.h>
#include <stdlib.h>
-int get_detection_layer_locations(detection_layer layer)
+detection_layer make_detection_layer(int batch, int inputs, int n, int side, int classes, int coords, int rescore)
{
- return layer.inputs / (layer.classes+layer.coords+layer.rescore);
-}
+ detection_layer l = {0};
+ l.type = DETECTION;
-int get_detection_layer_output_size(detection_layer layer)
-{
- return get_detection_layer_locations(layer)*(layer.classes+layer.coords);
-}
+ l.n = n;
+ l.batch = batch;
+ l.inputs = inputs;
+ l.classes = classes;
+ l.coords = coords;
+ l.rescore = rescore;
+ l.side = side;
+ l.w = side;
+ l.h = side;
+ assert(side*side*((1 + l.coords)*l.n + l.classes) == inputs);
+ l.cost = calloc(1, sizeof(float));
+ 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));
-detection_layer *make_detection_layer(int batch, int inputs, int classes, int coords, int rescore)
-{
- 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));
- #ifdef GPU
- layer->output_gpu = cuda_make_array(0, batch*outputs);
- layer->delta_gpu = cuda_make_array(0, batch*outputs);
- #endif
+ l.forward = forward_detection_layer;
+ l.backward = backward_detection_layer;
+#ifdef GPU
+ l.forward_gpu = forward_detection_layer_gpu;
+ l.backward_gpu = backward_detection_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
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 = l.side*l.side;
int i,j;
- for(i = 0; i < layer.batch*locations; ++i){
- int mask = (!truth || !truth[out_i + layer.classes - 1]);
- 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++];
+ memcpy(l.output, state.input, l.outputs*l.batch*sizeof(float));
+ //if(l.reorg) reorg(l.output, l.w*l.h, size*l.n, l.batch, 1);
+ 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(l.output + index + offset, l.classes, 1,
+ l.output + index + offset, 1);
+ }
}
- 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++];
+ }
+ 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 (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];
+ }
+
+ int best_index = -1;
+ float best_iou = 0;
+ float best_rmse = 20;
+
+ if (!is_obj){
+ 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 = 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 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;
+
+ if (l.sqrt){
+ out.w = out.w*out.w;
+ out.h = out.h*out.h;
+ }
+
+ float iou = box_iou(out, truth);
+ //iou = 0;
+ 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;
+ }
+ }
+ }
+
+ if(l.forced){
+ if(truth.w*truth.h < .1){
+ best_index = 1;
+ }else{
+ best_index = 0;
+ }
+ }
+ if(l.random && *(state.net.seen) < 64000){
+ best_index = rand()%l.n;
+ }
+
+ int box_index = index + locations*(l.classes + l.n) + (i*l.n + best_index) * l.coords;
+ int tbox_index = truth_index + 1 + l.classes;
+
+ box out = float_to_box(l.output + box_index);
+ out.x /= l.side;
+ out.y /= l.side;
+ if (l.sqrt) {
+ out.w = out.w*out.w;
+ out.h = out.h*out.h;
+ }
+ float iou = box_iou(out, truth);
+
+ //printf("%d,", best_index);
+ 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] = l.object_scale * (1.-l.output[p_index]);
+
+ if(l.rescore){
+ l.delta[p_index] = l.object_scale * (iou - l.output[p_index]);
+ }
+
+ 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-iou, 2);
+ avg_iou += iou;
+ ++count;
+ }
}
- //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(0){
+ float *costs = calloc(l.batch*locations*l.n, sizeof(float));
+ for (b = 0; b < l.batch; ++b) {
+ int index = b*l.inputs;
+ for (i = 0; i < locations; ++i) {
+ for (j = 0; j < l.n; ++j) {
+ int p_index = index + locations*l.classes + i*l.n + j;
+ costs[b*locations*l.n + i*l.n + j] = l.delta[p_index]*l.delta[p_index];
+ }
+ }
+ }
+ int indexes[100];
+ top_k(costs, l.batch*locations*l.n, 100, indexes);
+ float cutoff = costs[indexes[99]];
+ for (b = 0; b < l.batch; ++b) {
+ int index = b*l.inputs;
+ for (i = 0; i < locations; ++i) {
+ for (j = 0; j < l.n; ++j) {
+ int p_index = index + locations*l.classes + i*l.n + j;
+ if (l.delta[p_index]*l.delta[p_index] < cutoff) l.delta[p_index] = 0;
+ }
+ }
+ }
+ free(costs);
+ }
+
+
+ *(l.cost) = pow(mag_array(l.delta, l.outputs * l.batch), 2);
+
+
+ printf("Detection 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);
+ //if(l.reorg) reorg(l.delta, l.w*l.h, size*l.n, l.batch, 0);
}
}
-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 i,j;
- int in_i = 0;
- int out_i = 0;
- for(i = 0; i < layer.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++];
+ axpy_cpu(l.batch*l.inputs, 1, l.delta, 1, state.delta, 1);
+}
+
+void get_detection_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.side*l.side; ++i){
+ int row = i / l.side;
+ int col = i % l.side;
+ for(n = 0; n < l.n; ++n){
+ int index = i*l.n + n;
+ int p_index = l.side*l.side*l.classes + i*l.n + n;
+ float scale = predictions[p_index];
+ int box_index = l.side*l.side*(l.classes + l.n) + (i*l.n + n)*4;
+ boxes[index].x = (predictions[box_index + 0] + col) / l.side * w;
+ boxes[index].y = (predictions[box_index + 1] + row) / l.side * h;
+ boxes[index].w = pow(predictions[box_index + 2], (l.sqrt?2:1)) * w;
+ boxes[index].h = pow(predictions[box_index + 3], (l.sqrt?2:1)) * h;
+ for(j = 0; j < l.classes; ++j){
+ int class_index = i*l.classes;
+ float prob = scale*predictions[class_index+j];
+ probs[index][j] = (prob > thresh) ? prob : 0;
+ }
+ if(only_objectness){
+ probs[index][0] = scale;
+ }
}
-
- for(j = 0; j < layer.coords; ++j){
- delta[in_i++] = layer.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;
}
}
#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));
- 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.train){
+ copy_ongpu(l.batch*l.inputs, state.input, 1, l.output_gpu, 1);
+ return;
}
- 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);
+
+ float *in_cpu = calloc(l.batch*l.inputs, sizeof(float));
+ float *truth_cpu = 0;
+ if(state.truth){
+ 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 = 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*l.outputs);
+ cuda_push_array(l.delta_gpu, l.delta, l.batch*l.inputs);
+ 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);
-
- float *in_cpu = calloc(layer.batch*layer.inputs, sizeof(float));
- float *delta_cpu = calloc(layer.batch*layer.inputs, sizeof(float));
-
- 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);
-
- free(in_cpu);
- free(delta_cpu);
+ axpy_ongpu(l.batch*l.inputs, 1, l.delta_gpu, 1, state.delta, 1);
+ //copy_ongpu(l.batch*l.inputs, l.delta_gpu, 1, state.delta, 1);
}
#endif
--
Gitblit v1.10.0