From 846b3b4366c8d1a1ab7215db3ce4f4180ea53cb5 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Thu, 30 Jul 2015 16:19:14 -0700
Subject: [PATCH] Added COCO, fixed memory leaks

---
 Makefile                  |   2 +-
 src/box.c                 |  20 +++
 src/box.h                 |   1 +
 src/coco.c                | 297 ++++++++++++++++++++++++++++++++++++++
 src/convolutional_layer.c |  13 ++
 src/convolutional_layer.h |   2 +
 src/darknet.c             |  24 ++-
 src/detection.c           |  20 ---
 src/detection_layer.c     |   1 +
 src/image.c               |   1 +
 10 files changed, 359 insertions(+), 22 deletions(-)
 create mode 100644 src/coco.c

diff --git a/Makefile b/Makefile
index 1f8c84be..eff05bcb 100644
--- a/Makefile
+++ b/Makefile
@@ -34,7 +34,7 @@ CFLAGS+= -DGPU
 LDFLAGS+= -L/usr/local/cuda/lib64 -lcuda -lcudart -lcublas -lcurand
 endif
 
-OBJ=gemm.o utils.o cuda.o deconvolutional_layer.o convolutional_layer.o list.o image.o activations.o im2col.o col2im.o blas.o crop_layer.o dropout_layer.o maxpool_layer.o softmax_layer.o data.o matrix.o network.o connected_layer.o cost_layer.o parser.o option_list.o darknet.o detection_layer.o imagenet.o captcha.o detection.o route_layer.o writing.o box.o nightmare.o normalization_layer.o avgpool_layer.o
+OBJ=gemm.o utils.o cuda.o deconvolutional_layer.o convolutional_layer.o list.o image.o activations.o im2col.o col2im.o blas.o crop_layer.o dropout_layer.o maxpool_layer.o softmax_layer.o data.o matrix.o network.o connected_layer.o cost_layer.o parser.o option_list.o darknet.o detection_layer.o imagenet.o captcha.o detection.o route_layer.o writing.o box.o nightmare.o normalization_layer.o avgpool_layer.o coco.o
 ifeq ($(GPU), 1) 
 OBJ+=convolutional_kernels.o deconvolutional_kernels.o activation_kernels.o im2col_kernels.o col2im_kernels.o blas_kernels.o crop_layer_kernels.o dropout_layer_kernels.o maxpool_layer_kernels.o softmax_layer_kernels.o network_kernels.o avgpool_layer_kernels.o
 endif
diff --git a/src/box.c b/src/box.c
index cce56bd0..0518c050 100644
--- a/src/box.c
+++ b/src/box.c
@@ -211,3 +211,23 @@ dbox diou(box a, box b)
     return dd;
 }
 
+void do_nms(box *boxes, float **probs, int num_boxes, int classes, float thresh)
+{
+    int i, j, k;
+    for(i = 0; i < num_boxes*num_boxes; ++i){
+        int any = 0;
+        for(k = 0; k < classes; ++k) any = any || (probs[i][k] > 0);
+        if(!any) {
+            continue;
+        }
+        for(j = i+1; j < num_boxes*num_boxes; ++j){
+            if (box_iou(boxes[i], boxes[j]) > thresh){
+                for(k = 0; k < classes; ++k){
+                    if (probs[i][k] < probs[j][k]) probs[i][k] = 0;
+                    else probs[j][k] = 0;
+                }
+            }
+        }
+    }
+}
+
diff --git a/src/box.h b/src/box.h
index e3831d8e..998f58a2 100644
--- a/src/box.h
+++ b/src/box.h
@@ -11,5 +11,6 @@ typedef struct{
 
 float box_iou(box a, box b);
 dbox diou(box a, box b);
+void do_nms(box *boxes, float **probs, int num_boxes, int classes, float thresh);
 
 #endif
diff --git a/src/coco.c b/src/coco.c
new file mode 100644
index 00000000..af76eccb
--- /dev/null
+++ b/src/coco.c
@@ -0,0 +1,297 @@
+#include "network.h"
+#include "detection_layer.h"
+#include "cost_layer.h"
+#include "utils.h"
+#include "parser.h"
+#include "box.h"
+
+
+char *coco_classes[] = {"person","bicycle","car","motorcycle","airplane","bus","train","truck","boat","traffic light","fire hydrant","stop sign","parking meter","bench","bird","cat","dog","horse","sheep","cow","elephant","bear","zebra","giraffe","backpack","umbrella","handbag","tie","suitcase","frisbee","skis","snowboard","sports ball","kite","baseball bat","baseball glove","skateboard","surfboard","tennis racket","bottle","wine glass","cup","fork","knife","spoon","bowl","banana","apple","sandwich","orange","broccoli","carrot","hot dog","pizza","donut","cake","chair","couch","potted plant","bed","dining table","toilet","tv","laptop","mouse","remote","keyboard","cell phone","microwave","oven","toaster","sink","refrigerator","book","clock","vase","scissors","teddy bear","hair drier","toothbrush"};
+
+void draw_coco(image im, float *box, int side, int objectness, char *label)
+{
+    int classes = 80;
+    int elems = 4+classes+objectness;
+    int j;
+    int r, c;
+
+    for(r = 0; r < side; ++r){
+        for(c = 0; c < side; ++c){
+            j = (r*side + c) * elems;
+            float scale = 1;
+            if(objectness) scale = 1 - box[j++];
+            int class = max_index(box+j, classes);
+            if(scale * box[j+class] > 0.2){
+                int width = box[j+class]*5 + 1;
+                printf("%f %s\n", scale * box[j+class], coco_classes[class]);
+                float red = get_color(0,class,classes);
+                float green = get_color(1,class,classes);
+                float blue = get_color(2,class,classes);
+
+                j += classes;
+                float x = box[j+0];
+                float y = box[j+1];
+                x = (x+c)/side;
+                y = (y+r)/side;
+                float w = box[j+2]; //*maxwidth;
+                float h = box[j+3]; //*maxheight;
+                h = h*h;
+                w = w*w;
+
+                int left  = (x-w/2)*im.w;
+                int right = (x+w/2)*im.w;
+                int top   = (y-h/2)*im.h;
+                int bot   = (y+h/2)*im.h;
+                draw_box_width(im, left, top, right, bot, width, red, green, blue);
+            }
+        }
+    }
+    show_image(im, label);
+}
+
+void train_coco(char *cfgfile, char *weightfile)
+{
+    char *train_images = "/home/pjreddie/data/coco/train.txt";
+    char *backup_directory = "/home/pjreddie/backup/";
+    srand(time(0));
+    data_seed = time(0);
+    char *base = basecfg(cfgfile);
+    printf("%s\n", base);
+    float avg_loss = -1;
+    network net = parse_network_cfg(cfgfile);
+    if(weightfile){
+        load_weights(&net, weightfile);
+    }
+    detection_layer layer = get_network_detection_layer(net);
+    printf("Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay);
+    int imgs = 128;
+    int i = net.seen/imgs;
+    data train, buffer;
+
+    int classes = layer.classes;
+    int background = layer.objectness;
+    int side = sqrt(get_detection_layer_locations(layer));
+
+    char **paths;
+    list *plist = get_paths(train_images);
+    int N = plist->size;
+
+    paths = (char **)list_to_array(plist);
+    pthread_t load_thread = load_data_detection_thread(imgs, paths, plist->size, classes, net.w, net.h, side, side, background, &buffer);
+    clock_t time;
+    while(i*imgs < N*120){
+        i += 1;
+        time=clock();
+        pthread_join(load_thread, 0);
+        train = buffer;
+        load_thread = load_data_detection_thread(imgs, paths, plist->size, classes, net.w, net.h, side, side, background, &buffer);
+
+        printf("Loaded: %lf seconds\n", sec(clock()-time));
+
+        /*
+           image im = float_to_image(net.w, net.h, 3, train.X.vals[114]);
+           image copy = copy_image(im);
+           draw_coco(copy, train.y.vals[114], 7, layer.objectness, "truth");
+           cvWaitKey(0);
+           free_image(copy);
+         */
+
+        time=clock();
+        float loss = train_network(net, train);
+        net.seen += imgs;
+        if (avg_loss < 0) avg_loss = loss;
+        avg_loss = avg_loss*.9 + loss*.1;
+
+        printf("%d: %f, %f avg, %lf seconds, %d images\n", i, loss, avg_loss, sec(clock()-time), i*imgs);
+        if((i-1)*imgs <= 80*N && i*imgs > N*80){
+            fprintf(stderr, "First stage done.\n");
+            char buff[256];
+            sprintf(buff, "%s/%s_first_stage.weights", backup_directory, base);
+            save_weights(net, buff);
+            return;
+        }
+        if(i%1000==0 || 1){
+            char buff[256];
+            sprintf(buff, "%s/%s_%d.weights", backup_directory, base, i);
+            save_weights(net, buff);
+        }
+        free_data(train);
+        return;
+    }
+    char buff[256];
+    sprintf(buff, "%s/%s_final.weights", backup_directory, base);
+    save_weights(net, buff);
+}
+
+void convert_cocos(float *predictions, int classes, int objectness, int background, int num_boxes, int w, int h, float thresh, float **probs, box *boxes)
+{
+    int i,j;
+    int per_box = 4+classes+(background || objectness);
+    for (i = 0; i < num_boxes*num_boxes; ++i){
+        float scale = 1;
+        if(objectness) scale = 1-predictions[i*per_box];
+        int offset = i*per_box+(background||objectness);
+        for(j = 0; j < classes; ++j){
+            float prob = scale*predictions[offset+j];
+            probs[i][j] = (prob > thresh) ? prob : 0;
+        }
+        int row = i / num_boxes;
+        int col = i % num_boxes;
+        offset += classes;
+        boxes[i].x = (predictions[offset + 0] + col) / num_boxes * w;
+        boxes[i].y = (predictions[offset + 1] + row) / num_boxes * h;
+        boxes[i].w = pow(predictions[offset + 2], 2) * w;
+        boxes[i].h = pow(predictions[offset + 3], 2) * h;
+    }
+}
+
+void print_cocos(FILE **fps, char *id, box *boxes, float **probs, int num_boxes, int classes, int w, int h)
+{
+    int i, j;
+    for(i = 0; i < num_boxes*num_boxes; ++i){
+        float xmin = boxes[i].x - boxes[i].w/2.;
+        float xmax = boxes[i].x + boxes[i].w/2.;
+        float ymin = boxes[i].y - boxes[i].h/2.;
+        float ymax = boxes[i].y + boxes[i].h/2.;
+
+        if (xmin < 0) xmin = 0;
+        if (ymin < 0) ymin = 0;
+        if (xmax > w) xmax = w;
+        if (ymax > h) ymax = h;
+
+        for(j = 0; j < classes; ++j){
+            if (probs[i][j]) fprintf(fps[j], "%s %f %f %f %f %f\n", id, probs[i][j],
+                    xmin, ymin, xmax, ymax);
+        }
+    }
+}
+
+void validate_coco(char *cfgfile, char *weightfile)
+{
+    network net = parse_network_cfg(cfgfile);
+    if(weightfile){
+        load_weights(&net, weightfile);
+    }
+    set_batch_network(&net, 1);
+    detection_layer layer = get_network_detection_layer(net);
+    fprintf(stderr, "Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay);
+    srand(time(0));
+
+    char *base = "results/comp4_det_test_";
+    list *plist = get_paths("data/voc.2012test.list");
+    char **paths = (char **)list_to_array(plist);
+
+    int classes = layer.classes;
+    int objectness = layer.objectness;
+    int background = layer.background;
+    int num_boxes = sqrt(get_detection_layer_locations(layer));
+
+    int j;
+    FILE **fps = calloc(classes, sizeof(FILE *));
+    for(j = 0; j < classes; ++j){
+        char buff[1024];
+        snprintf(buff, 1024, "%s%s.txt", base, coco_classes[j]);
+        fps[j] = fopen(buff, "w");
+    }
+    box *boxes = calloc(num_boxes*num_boxes, sizeof(box));
+    float **probs = calloc(num_boxes*num_boxes, sizeof(float *));
+    for(j = 0; j < num_boxes*num_boxes; ++j) probs[j] = calloc(classes, sizeof(float *));
+
+    int m = plist->size;
+    int i=0;
+    int t;
+
+    float thresh = .001;
+    int nms = 1;
+    float iou_thresh = .5;
+
+    int nthreads = 8;
+    image *val = calloc(nthreads, sizeof(image));
+    image *val_resized = calloc(nthreads, sizeof(image));
+    image *buf = calloc(nthreads, sizeof(image));
+    image *buf_resized = calloc(nthreads, sizeof(image));
+    pthread_t *thr = calloc(nthreads, sizeof(pthread_t));
+    for(t = 0; t < nthreads; ++t){
+        thr[t] = load_image_thread(paths[i+t], &buf[t], &buf_resized[t], net.w, net.h);
+    }
+    time_t start = time(0);
+    for(i = nthreads; i < m+nthreads; i += nthreads){
+        fprintf(stderr, "%d\n", i);
+        for(t = 0; t < nthreads && i+t-nthreads < m; ++t){
+            pthread_join(thr[t], 0);
+            val[t] = buf[t];
+            val_resized[t] = buf_resized[t];
+        }
+        for(t = 0; t < nthreads && i+t < m; ++t){
+            thr[t] = load_image_thread(paths[i+t], &buf[t], &buf_resized[t], net.w, net.h);
+        }
+        for(t = 0; t < nthreads && i+t-nthreads < m; ++t){
+            char *path = paths[i+t-nthreads];
+            char *id = basecfg(path);
+            float *X = val_resized[t].data;
+            float *predictions = network_predict(net, X);
+            int w = val[t].w;
+            int h = val[t].h;
+            convert_cocos(predictions, classes, objectness, background, num_boxes, w, h, thresh, probs, boxes);
+            if (nms) do_nms(boxes, probs, num_boxes, classes, iou_thresh);
+            print_cocos(fps, id, boxes, probs, num_boxes, classes, w, h);
+            free(id);
+            free_image(val[t]);
+            free_image(val_resized[t]);
+        }
+    }
+    fprintf(stderr, "Total Detection Time: %f Seconds\n", (double)(time(0) - start));
+}
+
+void test_coco(char *cfgfile, char *weightfile, char *filename)
+{
+
+    network net = parse_network_cfg(cfgfile);
+    if(weightfile){
+        load_weights(&net, weightfile);
+    }
+    detection_layer layer = get_network_detection_layer(net);
+    set_batch_network(&net, 1);
+    srand(2222222);
+    clock_t time;
+    char input[256];
+    while(1){
+        if(filename){
+            strncpy(input, filename, 256);
+        } else {
+            printf("Enter Image Path: ");
+            fflush(stdout);
+            fgets(input, 256, stdin);
+            strtok(input, "\n");
+        }
+        image im = load_image_color(input,0,0);
+        image sized = resize_image(im, net.w, net.h);
+        float *X = sized.data;
+        time=clock();
+        float *predictions = network_predict(net, X);
+        printf("%s: Predicted in %f seconds.\n", input, sec(clock()-time));
+        draw_coco(im, predictions, 7, layer.objectness, "predictions");
+        free_image(im);
+        free_image(sized);
+#ifdef OPENCV
+        cvWaitKey(0);
+        cvDestroyAllWindows();
+#endif
+        if (filename) break;
+    }
+}
+
+void run_coco(int argc, char **argv)
+{
+    if(argc < 4){
+        fprintf(stderr, "usage: %s %s [train/test/valid] [cfg] [weights (optional)]\n", argv[0], argv[1]);
+        return;
+    }
+
+    char *cfg = argv[3];
+    char *weights = (argc > 4) ? argv[4] : 0;
+    char *filename = (argc > 5) ? argv[5]: 0;
+    if(0==strcmp(argv[2], "test")) test_coco(cfg, weights, filename);
+    else if(0==strcmp(argv[2], "train")) train_coco(cfg, weights);
+    else if(0==strcmp(argv[2], "valid")) validate_coco(cfg, weights);
+}
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index 378e23fa..7dcf5a43 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -242,6 +242,19 @@ void rgbgr_filters(convolutional_layer l)
     }
 }
 
+void rescale_filters(convolutional_layer l, float scale, float trans)
+{
+    int i;
+    for(i = 0; i < l.n; ++i){
+        image im = get_convolutional_filter(l, i);
+        if (im.c == 3) {
+            scale_image(im, scale);
+            float sum = sum_array(im.data, im.w*im.h*im.c);
+            l.biases[i] += sum*trans;
+        }
+    }
+}
+
 image *get_filters(convolutional_layer l)
 {
     image *filters = calloc(l.n, sizeof(image));
diff --git a/src/convolutional_layer.h b/src/convolutional_layer.h
index 3954f8a9..7452c3c5 100644
--- a/src/convolutional_layer.h
+++ b/src/convolutional_layer.h
@@ -38,6 +38,8 @@ image get_convolutional_filter(convolutional_layer layer, int i);
 
 int convolutional_out_height(convolutional_layer layer);
 int convolutional_out_width(convolutional_layer layer);
+void rescale_filters(convolutional_layer l, float scale, float trans);
+void rgbgr_filters(convolutional_layer l);
 
 #endif
 
diff --git a/src/darknet.c b/src/darknet.c
index 321b5a9f..d7fb1f5c 100644
--- a/src/darknet.c
+++ b/src/darknet.c
@@ -11,6 +11,7 @@
 
 extern void run_imagenet(int argc, char **argv);
 extern void run_detection(int argc, char **argv);
+extern void run_coco(int argc, char **argv);
 extern void run_writing(int argc, char **argv);
 extern void run_captcha(int argc, char **argv);
 extern void run_nightmare(int argc, char **argv);
@@ -40,7 +41,24 @@ void partial(char *cfgfile, char *weightfile, char *outfile, int max)
 }
 
 #include "convolutional_layer.h"
-void rgbgr_filters(convolutional_layer l);
+void rescale_net(char *cfgfile, char *weightfile, char *outfile)
+{
+    gpu_index = -1;
+    network net = parse_network_cfg(cfgfile);
+    if(weightfile){
+        load_weights(&net, weightfile);
+    }
+    int i;
+    for(i = 0; i < net.n; ++i){
+        layer l = net.layers[i];
+        if(l.type == CONVOLUTIONAL){
+            rescale_filters(l, 2, -.5);
+            break;
+        }
+    }
+    save_weights(net, outfile);
+}
+
 void rgbgr_net(char *cfgfile, char *weightfile, char *outfile)
 {
     gpu_index = -1;
@@ -95,6 +113,8 @@ int main(int argc, char **argv)
         run_imagenet(argc, argv);
     } else if (0 == strcmp(argv[1], "detection")){
         run_detection(argc, argv);
+    } else if (0 == strcmp(argv[1], "coco")){
+        run_coco(argc, argv);
     } else if (0 == strcmp(argv[1], "writing")){
         run_writing(argc, argv);
     } else if (0 == strcmp(argv[1], "test")){
@@ -107,6 +127,8 @@ int main(int argc, char **argv)
         change_rate(argv[2], atof(argv[3]), (argc > 4) ? atof(argv[4]) : 0);
     } else if (0 == strcmp(argv[1], "rgbgr")){
         rgbgr_net(argv[2], argv[3], argv[4]);
+    } else if (0 == strcmp(argv[1], "rescale")){
+        rescale_net(argv[2], argv[3], argv[4]);
     } else if (0 == strcmp(argv[1], "partial")){
         partial(argv[2], argv[3], argv[4], atoi(argv[5]));
     } else if (0 == strcmp(argv[1], "visualize")){
diff --git a/src/detection.c b/src/detection.c
index 615ad6d2..55c75dee 100644
--- a/src/detection.c
+++ b/src/detection.c
@@ -143,26 +143,6 @@ void convert_detections(float *predictions, int classes, int objectness, int bac
     }
 }
 
-void do_nms(box *boxes, float **probs, int num_boxes, int classes, float thresh)
-{
-    int i, j, k;
-    for(i = 0; i < num_boxes*num_boxes; ++i){
-        int any = 0;
-        for(k = 0; k < classes; ++k) any = any || (probs[i][k] > 0);
-        if(!any) {
-            continue;
-        }
-        for(j = i+1; j < num_boxes*num_boxes; ++j){
-            if (box_iou(boxes[i], boxes[j]) > thresh){
-                for(k = 0; k < classes; ++k){
-                    if (probs[i][k] < probs[j][k]) probs[i][k] = 0;
-                    else probs[j][k] = 0;
-                }
-            }
-        }
-    }
-}
-
 void print_detections(FILE **fps, char *id, box *boxes, float **probs, int num_boxes, int classes, int w, int h)
 {
     int i, j;
diff --git a/src/detection_layer.c b/src/detection_layer.c
index 6a25819b..e48b8b37 100644
--- a/src/detection_layer.c
+++ b/src/detection_layer.c
@@ -204,6 +204,7 @@ void backward_detection_layer_gpu(detection_layer l, network_state state)
     backward_detection_layer(l, cpu_state);
     cuda_push_array(state.delta, delta_cpu, l.batch*l.inputs);
 
+    if (truth_cpu) free(truth_cpu);
     free(in_cpu);
     free(delta_cpu);
 }
diff --git a/src/image.c b/src/image.c
index 5db93f16..92c90665 100644
--- a/src/image.c
+++ b/src/image.c
@@ -249,6 +249,7 @@ void save_image(image im, char *name)
         }
     }
     int success = stbi_write_png(buff, im.w, im.h, im.c, data, im.w*im.c);
+    free(data);
     if(!success) fprintf(stderr, "Failed to write image %s\n", buff);
 }