From 3ebcc647b651a4a3c717eff2a3087127e5707e0c Mon Sep 17 00:00:00 2001
From: AlexeyAB <alexeyab84@gmail.com>
Date: Mon, 14 May 2018 09:20:38 +0000
Subject: [PATCH] Fixed network resizing (random=1) for non-square networks
---
src/detector.c | 73 +++++++++++++++++++++++++++++-------
1 files changed, 58 insertions(+), 15 deletions(-)
diff --git a/src/detector.c b/src/detector.c
index ea5b49d..2b54a4a 100644
--- a/src/detector.c
+++ b/src/detector.c
@@ -1,3 +1,8 @@
+#ifdef _DEBUG
+#include <stdlib.h>
+#include <crtdbg.h>
+#endif
+
#include "network.h"
#include "region_layer.h"
#include "cost_layer.h"
@@ -61,6 +66,11 @@
srand(time(0));
network net = nets[0];
+ if ((net.batch * net.subdivisions) == 1) {
+ printf("\n Error: You set incorrect value batch=1 for Training! You should set batch=64 subdivision=64 \n");
+ getchar();
+ }
+
int imgs = net.batch * net.subdivisions * ngpus;
printf("Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay);
data train, buffer;
@@ -116,12 +126,16 @@
while(get_current_batch(net) < net.max_batches){
if(l.random && count++%10 == 0){
printf("Resizing\n");
- int dim = (rand() % 12 + (init_w/32 - 5)) * 32; // +-160
- //if (get_current_batch(net)+100 > net.max_batches) dim = 544;
+ //int dim = (rand() % 12 + (init_w/32 - 5)) * 32; // +-160
//int dim = (rand() % 4 + 16) * 32;
- printf("%d\n", dim);
- args.w = dim;
- args.h = dim;
+ //if (get_current_batch(net)+100 > net.max_batches) dim = 544;
+ int random_val = rand() % 12;
+ int dim_w = (random_val + (init_w / 32 - 5)) * 32; // +-160
+ int dim_h = (random_val + (init_h / 32 - 5)) * 32; // +-160
+
+ printf("%d x %d \n", dim_w, dim_h);
+ args.w = dim_w;
+ args.h = dim_h;
pthread_join(load_thread, 0);
train = buffer;
@@ -129,7 +143,7 @@
load_thread = load_data(args);
for(i = 0; i < ngpus; ++i){
- resize_network(nets + i, dim, dim);
+ resize_network(nets + i, dim_w, dim_h);
}
net = nets[0];
}
@@ -645,6 +659,8 @@
truth_dif = read_boxes(labelpath_dif, &num_labels_dif);
}
+ const int checkpoint_detections_count = detections_count;
+
for (i = 0; i < nboxes; ++i) {
int class_id;
@@ -695,7 +711,13 @@
// calc avg IoU, true-positives, false-positives for required Threshold
if (prob > thresh_calc_avg_iou) {
- if (truth_index > -1) {
+ int z, found = 0;
+ for (z = checkpoint_detections_count; z < detections_count-1; ++z)
+ if (detections[z].unique_truth_index == truth_index) {
+ found = 1; break;
+ }
+
+ if(truth_index > -1 && found == 0) {
avg_iou += max_iou;
++tp_for_thresh;
}
@@ -715,7 +737,8 @@
}
}
- avg_iou = avg_iou / (tp_for_thresh + fp_for_thresh);
+ if((tp_for_thresh + fp_for_thresh) > 0)
+ avg_iou = avg_iou / (tp_for_thresh + fp_for_thresh);
// SORT(detections)
@@ -1027,7 +1050,8 @@
}
#endif // OPENCV
-void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh, float hier_thresh, int dont_show)
+void test_detector(char *datacfg, char *cfgfile, char *weightfile, char *filename, float thresh,
+ float hier_thresh, int dont_show, int ext_output)
{
list *options = read_data_cfg(datacfg);
char *name_list = option_find_str(options, "names", "data/names.list");
@@ -1060,8 +1084,8 @@
}
image im = load_image_color(input,0,0);
int letterbox = 0;
- image sized = resize_image(im, net.w, net.h);
- //image sized = letterbox_image(im, net.w, net.h); letterbox = 1;
+ //image sized = resize_image(im, net.w, net.h);
+ image sized = letterbox_image(im, net.w, net.h); letterbox = 1;
layer l = net.layers[net.n-1];
//box *boxes = calloc(l.w*l.h*l.n, sizeof(box));
@@ -1070,8 +1094,8 @@
float *X = sized.data;
time= what_time_is_it_now();
- //network_predict(net, X);
- network_predict_image(&net, im); letterbox = 1;
+ network_predict(net, X);
+ //network_predict_image(&net, im); letterbox = 1;
printf("%s: Predicted in %f seconds.\n", input, (what_time_is_it_now()-time));
//get_region_boxes(l, 1, 1, thresh, probs, boxes, 0, 0);
// if (nms) do_nms_sort_v2(boxes, probs, l.w*l.h*l.n, l.classes, nms);
@@ -1079,7 +1103,7 @@
int nboxes = 0;
detection *dets = get_network_boxes(&net, im.w, im.h, thresh, hier_thresh, 0, 1, &nboxes, letterbox);
if (nms) do_nms_sort(dets, nboxes, l.classes, nms);
- draw_detections_v3(im, dets, nboxes, thresh, names, alphabet, l.classes);
+ draw_detections_v3(im, dets, nboxes, thresh, names, alphabet, l.classes, ext_output);
free_detections(dets, nboxes);
save_image(im, "predictions");
if (!dont_show) {
@@ -1098,6 +1122,22 @@
#endif
if (filename) break;
}
+
+ // free memory
+ free_ptrs(names, net.layers[net.n - 1].classes);
+ free_list(options);
+
+ int i;
+ const int nsize = 8;
+ for (j = 0; j < nsize; ++j) {
+ for (i = 32; i < 127; ++i) {
+ free_image(alphabet[j][i]);
+ }
+ free(alphabet[j]);
+ }
+ free(alphabet);
+
+ free_network(net);
}
void run_detector(int argc, char **argv)
@@ -1115,6 +1155,9 @@
int num_of_clusters = find_int_arg(argc, argv, "-num_of_clusters", 5);
int width = find_int_arg(argc, argv, "-width", -1);
int height = find_int_arg(argc, argv, "-height", -1);
+ // extended output in test mode (output of rect bound coords)
+ // and for recall mode (extended output table-like format with results for best_class fit)
+ int ext_output = find_arg(argc, argv, "-ext_output");
if(argc < 4){
fprintf(stderr, "usage: %s %s [train/test/valid] [cfg] [weights (optional)]\n", argv[0], argv[1]);
return;
@@ -1151,7 +1194,7 @@
if(strlen(weights) > 0)
if (weights[strlen(weights) - 1] == 0x0d) weights[strlen(weights) - 1] = 0;
char *filename = (argc > 6) ? argv[6]: 0;
- if(0==strcmp(argv[2], "test")) test_detector(datacfg, cfg, weights, filename, thresh, hier_thresh, dont_show);
+ if(0==strcmp(argv[2], "test")) test_detector(datacfg, cfg, weights, filename, thresh, hier_thresh, dont_show, ext_output);
else if(0==strcmp(argv[2], "train")) train_detector(datacfg, cfg, weights, gpus, ngpus, clear, dont_show);
else if(0==strcmp(argv[2], "valid")) validate_detector(datacfg, cfg, weights, outfile);
else if(0==strcmp(argv[2], "recall")) validate_detector_recall(datacfg, cfg, weights);
--
Gitblit v1.10.0