Experimental reinforcement learning.

AlexeyAB

2018-05-22 032acd408acd119a5bd5c132765fcdd6caedea6b

Experimental reinforcement learning.

 src/detector.c

@@ -542,6 +542,7 @@
    char *mapf = option_find_str(options, "map", 0);
    int *map = 0;
    if (mapf) map = read_map(mapf);
    FILE* reinforcement_fd = NULL;

    network net = parse_network_cfg_custom(cfgfile, 1); // set batch=1
    if (weightfile) {
@@ -720,9 +721,18 @@
                    }
                }
            }
			
				
            unique_truth_count += num_labels;

            //static int previous_errors = 0;
            //int total_errors = fp_for_thresh + (unique_truth_count - tp_for_thresh);
            //int errors_in_this_image = total_errors - previous_errors;
            //previous_errors = total_errors;
            //if(reinforcement_fd == NULL) reinforcement_fd = fopen("reinforcement.txt", "wb");
            //char buff[1000];
            //sprintf(buff, "%s\n", path);
            //if(errors_in_this_image > 0) fwrite(buff, sizeof(char), strlen(buff), reinforcement_fd);

            free_detections(dets, nboxes);
            free(id);
            free_image(val[t]);
@@ -843,6 +853,7 @@
    free(truth_classes_count);

    fprintf(stderr, "Total Detection Time: %f Seconds\n", (double)(time(0) - start));
    if (reinforcement_fd != NULL) fclose(reinforcement_fd);
}

#ifdef OPENCV