idk, probably something changed

src/crop_layer.c

@@ -28,14 +28,19 @@ crop_layer *make_crop_layer(int batch, int h, int w, int c, int crop_height, int
     return layer;
 }
 
-void forward_crop_layer(const crop_layer layer, float *input)
+void forward_crop_layer(const crop_layer layer, int train, float *input)
 {
     int i,j,c,b,row,col;
     int index;
     int count = 0;
     int flip = (layer.flip && rand()%2);
-    int dh = rand()%(layer.h - layer.crop_height);
+    int dh = rand()%(layer.h - layer.crop_height + 1);
-    int dw = rand()%(layer.w - layer.crop_width);
+    int dw = rand()%(layer.w - layer.crop_width + 1);
+    if(!train){
+        flip = 0;
+        dh = (layer.h - layer.crop_height)/2;
+        dw = (layer.w - layer.crop_width)/2;
+    }
     for(b = 0; b < layer.batch; ++b){
         for(c = 0; c < layer.c; ++c){
             for(i = 0; i < layer.crop_height; ++i){

src/crop_layer.h

@@ -17,10 +17,10 @@ typedef struct {
 
 image get_crop_image(crop_layer layer);
 crop_layer *make_crop_layer(int batch, int h, int w, int c, int crop_height, int crop_width, int flip);
-void forward_crop_layer(const crop_layer layer, float *input);
+void forward_crop_layer(const crop_layer layer, int train, float *input);
 
 #ifdef GPU
-void forward_crop_layer_gpu(crop_layer layer, float *input);
+void forward_crop_layer_gpu(crop_layer layer, int train, float *input);
 #endif
 
 #endif

src/crop_layer_kernels.cu

@@ -24,11 +24,16 @@ __global__ void forward_crop_layer_kernel(float *input, int size, int c, int h,
     output[count] = input[index];
 }
 
-extern "C" void forward_crop_layer_gpu(crop_layer layer, float *input)
+extern "C" void forward_crop_layer_gpu(crop_layer layer, int train, float *input)
 {
     int flip = (layer.flip && rand()%2);
-    int dh = rand()%(layer.h - layer.crop_height);
+    int dh = rand()%(layer.h - layer.crop_height + 1);
-    int dw = rand()%(layer.w - layer.crop_width);
+    int dw = rand()%(layer.w - layer.crop_width + 1);
+    if(!train){
+        flip = 0;
+        dh = (layer.h - layer.crop_height)/2;
+        dw = (layer.w - layer.crop_width)/2;
+    }
     int size = layer.batch*layer.c*layer.crop_width*layer.crop_height;
 
     dim3 dimBlock(BLOCK, 1, 1);

src/cuda.c

@@ -1,9 +1,12 @@
+int gpu_index = 0;
+
+#ifdef GPU
+
 #include "cuda.h"
 #include "utils.h"
 #include "blas.h"
 #include <stdlib.h>
 
-int gpu_index = 0;
 
 void check_error(cudaError_t status)
 {
@@ -96,4 +99,4 @@ void cuda_pull_array(float *x_gpu, float *x, int n)
     check_error(status);
 }
 
-
+#endif

src/cuda.h

@@ -1,13 +1,15 @@
 #ifndef CUDA_H
 #define CUDA_H
 
+extern int gpu_index;
+
+#ifdef GPU
+
 #define BLOCK 256
 
 #include "cuda_runtime.h"
 #include "cublas_v2.h"
 
-extern int gpu_index;
-
 void check_error(cudaError_t status);
 cublasHandle_t blas_handle();
 float *cuda_make_array(float *x, int n);
@@ -19,3 +21,4 @@ float cuda_compare(float *x_gpu, float *x, int n, char *s);
 dim3 cuda_gridsize(size_t n);
 
 #endif
+#endif

src/darknet.c

@@ -209,13 +209,12 @@ void train_imagenet_distributed(char *address)
 void train_imagenet(char *cfgfile)
 {
     float avg_loss = 1;
-    //network net = parse_network_cfg("/home/pjreddie/imagenet_backup/alexnet_1270.cfg");
     srand(time(0));
     network net = parse_network_cfg(cfgfile);
     //test_learn_bias(*(convolutional_layer *)net.layers[1]);
     //set_learning_network(&net, net.learning_rate, 0, net.decay);
     printf("Learning Rate: %g, Momentum: %g, Decay: %g\n", net.learning_rate, net.momentum, net.decay);
-    int imgs = 3072;
+    int imgs = 1024;
     int i = net.seen/imgs;
     char **labels = get_labels("/home/pjreddie/data/imagenet/cls.labels.list");
     list *plist = get_paths("/data/imagenet/cls.train.list");
@@ -231,9 +230,6 @@ void train_imagenet(char *cfgfile)
         time=clock();
         pthread_join(load_thread, 0);
         train = buffer;
-        //normalize_data_rows(train);
-        //translate_data_rows(train, -128);
-        //scale_data_rows(train, 1./128);
         load_thread = load_data_thread(paths, imgs, plist->size, labels, 1000, 256, 256, &buffer);
         printf("Loaded: %lf seconds\n", sec(clock()-time));
         time=clock();
@@ -244,7 +240,7 @@ void train_imagenet(char *cfgfile)
         free_data(train);
         if(i%100==0){
             char buff[256];
-            sprintf(buff, "/home/pjreddie/imagenet_backup/alexnet_%d.cfg", i);
+            sprintf(buff, "/home/pjreddie/imagenet_backup/vgg_%d.cfg", i);
             save_network(net, buff);
         }
     }
@@ -347,10 +343,28 @@ void test_init(char *cfgfile)
     }
     free_image(im);
 }
-
+void test_dog(char *cfgfile)
-void test_imagenet()
 {
-    network net = parse_network_cfg("cfg/imagenet_test.cfg");
+    image im = load_image_color("data/dog.jpg", 224, 224);
+    translate_image(im, -128);
+    print_image(im);
+    float *X = im.data;
+    network net = parse_network_cfg(cfgfile);
+    set_batch_network(&net, 1);
+    float *predictions = network_predict(net, X);
+    image crop = get_network_image_layer(net, 0);
+    //show_image(crop, "cropped");
+   // print_image(crop);
+    //show_image(im, "orig");
+    float * inter = get_network_output(net);
+    pm(1000, 1, inter);
+    //cvWaitKey(0);
+}
+
+void test_imagenet(char *cfgfile)
+{
+    network net = parse_network_cfg(cfgfile);
+    set_batch_network(&net, 1);
     //imgs=1;
     srand(2222222);
     int i = 0;
@@ -362,7 +376,8 @@ void test_imagenet()
         fgets(filename, 256, stdin);
         strtok(filename, "\n");
         image im = load_image_color(filename, 256, 256);
-        z_normalize_image(im);
+        translate_image(im, -128);
+        //scale_image(im, 1/128.);
         printf("%d %d %d\n", im.h, im.w, im.c);
         float *X = im.data;
         time=clock();
@@ -472,28 +487,28 @@ void train_nist(char *cfgfile)
 }
 
 /*
-void train_nist_distributed(char *address)
+   void train_nist_distributed(char *address)
-{
+   {
    srand(time(0));
    network net = parse_network_cfg("cfg/nist.client");
    data train = load_categorical_data_csv("data/mnist/mnist_train.csv", 0, 10);
-    //data test = load_categorical_data_csv("data/mnist/mnist_test.csv",0,10);
+//data test = load_categorical_data_csv("data/mnist/mnist_test.csv",0,10);
-    normalize_data_rows(train);
+normalize_data_rows(train);
-    //normalize_data_rows(test);
+//normalize_data_rows(test);
-    int count = 0;
+int count = 0;
-    int iters = 50000/net.batch;
+int iters = 50000/net.batch;
-    iters = 1000/net.batch + 1;
+iters = 1000/net.batch + 1;
-    while(++count <= 2000){
+while(++count <= 2000){
-        clock_t start = clock(), end;
+clock_t start = clock(), end;
-        float loss = train_network_sgd(net, train, iters);
+float loss = train_network_sgd(net, train, iters);
-        client_update(net, address);
+client_update(net, address);
-        end = clock();
+end = clock();
-        //float test_acc = network_accuracy_gpu(net, test);
+//float test_acc = network_accuracy_gpu(net, test);
-        //float test_acc = 0;
+//float test_acc = 0;
-        printf("%d: Loss: %f, Time: %lf seconds\n", count, loss, (float)(end-start)/CLOCKS_PER_SEC);
+printf("%d: Loss: %f, Time: %lf seconds\n", count, loss, (float)(end-start)/CLOCKS_PER_SEC);
-    }
 }
-*/
+}
+ */
 
 void test_ensemble()
 {
@@ -535,7 +550,7 @@ void test_ensemble()
 void visualize_cat()
 {
     network net = parse_network_cfg("cfg/voc_imagenet.cfg");
-    image im = load_image("data/cat.png", 0, 0);
+    image im = load_image_color("data/cat.png", 0, 0);
     printf("Processing %dx%d image\n", im.h, im.w);
     resize_network(net, im.h, im.w, im.c);
     forward_network(net, im.data, 0, 0);
@@ -544,6 +559,7 @@ void visualize_cat()
     cvWaitKey(0);
 }
 
+#ifdef GPU
 void test_convolutional_layer()
 {
     network net = parse_network_cfg("cfg/nist_conv.cfg");
@@ -561,6 +577,7 @@ void test_convolutional_layer()
     bias_output_gpu(layer);
     cuda_compare(layer.output_gpu, layer.output, out_size, "biased output");
 }
+#endif
 
 void test_correct_nist()
 {
@@ -641,16 +658,16 @@ void test_correct_alexnet()
 }
 
 /*
-void run_server()
+   void run_server()
-{
+   {
    srand(time(0));
    network net = parse_network_cfg("cfg/net.cfg");
    set_batch_network(&net, 1);
    server_update(net);
-}
+   }
 
-void test_client()
+   void test_client()
-{
+   {
    network net = parse_network_cfg("cfg/alexnet.client");
    clock_t time=clock();
    client_update(net, "localhost");
@@ -660,8 +677,8 @@ void test_client()
    client_update(net, "localhost");
    printf("3\n");
    printf("Transfered: %lf seconds\n", sec(clock()-time));
-}
+   }
-*/
+ */
 
 void del_arg(int argc, char **argv, int index)
 {
@@ -713,7 +730,6 @@ int main(int argc, char **argv)
 
     if(0==strcmp(argv[1], "test_correct")) test_correct_alexnet();
     else if(0==strcmp(argv[1], "test_correct_nist")) test_correct_nist();
-    else if(0==strcmp(argv[1], "test")) test_imagenet();
     //else if(0==strcmp(argv[1], "server")) run_server();
 
 #ifdef GPU
@@ -725,6 +741,8 @@ int main(int argc, char **argv)
         return 0;
     }
     else if(0==strcmp(argv[1], "detection")) train_detection_net(argv[2]);
+    else if(0==strcmp(argv[1], "test")) test_imagenet(argv[2]);
+    else if(0==strcmp(argv[1], "dog")) test_dog(argv[2]);
     else if(0==strcmp(argv[1], "ctrain")) train_cifar10(argv[2]);
     else if(0==strcmp(argv[1], "nist")) train_nist(argv[2]);
     else if(0==strcmp(argv[1], "ctest")) test_cifar10(argv[2]);

src/data.c

@@ -239,7 +239,7 @@ void *load_in_thread(void *ptr)
 {
     struct load_args a = *(struct load_args*)ptr;
     *a.d = load_data(a.paths, a.n, a.m, a.labels, a.k, a.h, a.w);
-	translate_data_rows(*a.d, -144);
+	translate_data_rows(*a.d, -128);
 	scale_data_rows(*a.d, 1./128);
     free(ptr);
     return 0;

src/image.c

@@ -484,7 +484,7 @@ image load_image(char *filename, int h, int w)
         exit(0);
     }
     if(h && w ){
-        IplImage *resized = resizeImage(src, h, w, 1);
+        IplImage *resized = resizeImage(src, h, w, 0);
         cvReleaseImage(&src);
         src = resized;
     }
@@ -702,10 +702,21 @@ void back_convolve(image m, image kernel, int stride, int channel, image out, in
 
 void print_image(image m)
 {
-    int i;
+    int i, j, k;
-    for(i =0 ; i < m.h*m.w*m.c; ++i) printf("%lf, ", m.data[i]);
+    for(i =0 ; i < m.c; ++i){
+        for(j =0 ; j < m.h; ++j){
+            for(k = 0; k < m.w; ++k){
+                printf("%.2lf, ", m.data[i*m.h*m.w + j*m.w + k]);
+                if(k > 30) break;
+            }
+            printf("\n");
+            if(j > 30) break;
+        }
+        printf("\n");
+    }
     printf("\n");
 }
+
 image collapse_images_vert(image *ims, int n)
 {
     int color = 1;

src/network.c

@@ -75,7 +75,7 @@ void forward_network(network net, float *input, float *truth, int train)
         }
         else if(net.types[i] == CROP){
             crop_layer layer = *(crop_layer *)net.layers[i];
-            forward_crop_layer(layer, input);
+            forward_crop_layer(layer, train, input);
             input = layer.output;
         }
         else if(net.types[i] == COST){
@@ -536,6 +536,9 @@ image get_network_image_layer(network net, int i)
         normalization_layer layer = *(normalization_layer *)net.layers[i];
         return get_normalization_image(layer);
     }
+    else if(net.types[i] == DROPOUT){
+        return get_network_image_layer(net, i-1);
+    }
     else if(net.types[i] == CROP){
         crop_layer layer = *(crop_layer *)net.layers[i];
         return get_crop_image(layer);

src/network_kernels.cu

@@ -58,7 +58,7 @@ void forward_network_gpu(network net, float * input, float * truth, int train)
         }
         else if(net.types[i] == CROP){
             crop_layer layer = *(crop_layer *)net.layers[i];
-            forward_crop_layer_gpu(layer, input);
+            forward_crop_layer_gpu(layer, train, input);
             input = layer.output_gpu;
         }
         //printf("Forward %d %s %f\n", i, get_layer_string(net.types[i]), sec(clock() - time));

src/utils.c

@@ -11,6 +11,7 @@ void pm(int M, int N, float *A)
 {
     int i,j;
     for(i =0 ; i < M; ++i){
+        printf("%d ", i+1);
         for(j = 0; j < N; ++j){
             printf("%10.6f, ", A[i*N+j]);
         }