NIPS

Makefile

@@ -1,5 +1,7 @@
 GPU=1
+OPENCV=1
 DEBUG=0
+
 ARCH= -arch=sm_52
 
 VPATH=./src/
@@ -9,19 +11,26 @@ OBJDIR=./obj/
 CC=gcc
 NVCC=nvcc
 OPTS=-Ofast
-LDFLAGS=`pkg-config --libs opencv` -lm -pthread -lstdc++
-COMMON=`pkg-config --cflags opencv` -I/usr/local/cuda/include/
-CFLAGS=-Wall -Wfatal-errors
+LDFLAGS= -lm -pthread -lstdc++ 
+COMMON= -I/usr/local/cuda/include/ 
+CFLAGS=-Wall -Wfatal-errors 
 
 ifeq ($(DEBUG), 1) 
 OPTS=-O0 -g
 endif
 
 CFLAGS+=$(OPTS)
 
+ifeq ($(OPENCV), 1) 
+COMMON+= -DOPENCV
+CFLAGS+= -DOPENCV
+LDFLAGS+= `pkg-config --libs opencv` 
+COMMON+= `pkg-config --cflags opencv` 
+endif
+
 ifeq ($(GPU), 1) 
-COMMON+=-DGPU
-CFLAGS+=-DGPU
+COMMON+= -DGPU
+CFLAGS+= -DGPU
 LDFLAGS+= -L/usr/local/cuda/lib64 -lcuda -lcudart -lcublas -lcurand
 endif
 

cfg/rescore.cfg → cfg/yolo.cfg

@@ -1,6 +1,6 @@
 [net]
-batch=64
-subdivisions=4
+batch=1
+subdivisions=1
 height=448
 width=448
 channels=3
@@ -199,6 +199,7 @@ activation=logistic
 [detection]
 classes=20
 coords=4
-rescore=1
-nuisance = 0
+rescore=0
+joint=1
+objectness = 0
 background=0

src/captcha.c

@@ -62,7 +62,9 @@ void decode_captcha(char *cfgfile, char *weightfile)
         float *predictions = network_predict(net, X);
         image out  = float_to_image(300, 57, 1, predictions);
         show_image(out, "decoded");
+        #ifdef OPENCV
         cvWaitKey(0);
+        #endif
         free_image(im);
     }
 }

src/convolutional_layer.c

@@ -227,6 +227,15 @@ image get_convolutional_filter(convolutional_layer l, int i)
     return float_to_image(w,h,c,l.filters+i*h*w*c);
 }
 
+void rgbgr_filters(convolutional_layer l)
+{
+    int i;
+    for(i = 0; i < l.n; ++i){
+        image im = get_convolutional_filter(l, i);
+        if (im.c == 3) rgbgr_image(im);
+    }
+}
+
 image *get_filters(convolutional_layer l)
 {
     image *filters = calloc(l.n, sizeof(image));

src/crop_layer_kernels.cu

@@ -114,19 +114,19 @@ __global__ void levels_image_kernel(float *image, float *rand, int batch, int w,
 
     size_t offset = id * h * w * 3;
     image += offset;
-    float r = image[x + w*(y + h*2)];
+    float r = image[x + w*(y + h*0)];
     float g = image[x + w*(y + h*1)];
-    float b = image[x + w*(y + h*0)];
+    float b = image[x + w*(y + h*2)];
     float3 rgb = make_float3(r,g,b);
     if(train){
         float3 hsv = rgb_to_hsv_kernel(rgb);
         hsv.y *= saturation;
         hsv.z *= exposure;
         rgb = hsv_to_rgb_kernel(hsv);
     }
-    image[x + w*(y + h*2)] = rgb.x*scale + translate;
+    image[x + w*(y + h*0)] = rgb.x*scale + translate;
     image[x + w*(y + h*1)] = rgb.y*scale + translate;
-    image[x + w*(y + h*0)] = rgb.z*scale + translate;
+    image[x + w*(y + h*2)] = rgb.z*scale + translate;
 }
 
 __global__ void forward_crop_layer_kernel(float *input, float *rand, int size, int c, int h, int w, int crop_height, int crop_width, int train, int flip, float angle, float *output)

src/darknet.c

@@ -73,18 +73,40 @@ void partial(char *cfgfile, char *weightfile, char *outfile, int max)
     save_weights(net, outfile);
 }
 
+#include "convolutional_layer.h"
+void rgbgr_filters(convolutional_layer l);
+void rgbgr_net(char *cfgfile, char *weightfile, char *outfile)
+{
+    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){
+            rgbgr_filters(l);
+            break;
+        }
+    }
+    save_weights(net, outfile);
+}
+
 void visualize(char *cfgfile, char *weightfile)
 {
     network net = parse_network_cfg(cfgfile);
     if(weightfile){
         load_weights(&net, weightfile);
     }
     visualize_network(net);
+    #ifdef OPENCV
     cvWaitKey(0);
+    #endif
 }
 
 int main(int argc, char **argv)
 {
+    //test_resize("data/cat.png");
     //test_box();
     //test_convolutional_layer();
     if(argc < 2){
@@ -114,6 +136,8 @@ int main(int argc, char **argv)
         run_captcha(argc, argv);
     } else if (0 == strcmp(argv[1], "change")){
         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], "partial")){
         partial(argv[2], argv[3], argv[4], atoi(argv[5]));
     } else if (0 == strcmp(argv[1], "visualize")){

src/data.c

@@ -69,7 +69,7 @@ matrix load_image_paths_gray(char **paths, int n, int w, int h)
     X.cols = 0;
 
     for(i = 0; i < n; ++i){
-        image im = load_image(paths[i], w, h);
+        image im = load_image(paths[i], w, h, 1);
         X.vals[i] = im.data;
         X.cols = im.h*im.w*im.c;
     }