Extracting features from VOC with temp filters

src/convolutional_layer.c

@@ -3,11 +3,21 @@
 #include "mini_blas.h"
 #include <stdio.h>
 
+int convolutional_out_height(convolutional_layer layer)
+{
+    return (layer.h-layer.size)/layer.stride + 1;
+}
+
+int convolutional_out_width(convolutional_layer layer)
+{
+    return (layer.w-layer.size)/layer.stride + 1;
+}
+
 image get_convolutional_image(convolutional_layer layer)
 {
     int h,w,c;
-    h = layer.out_h;
-    w = layer.out_w;
+    h = convolutional_out_height(layer);
+    w = convolutional_out_width(layer);
     c = layer.n;
     return float_to_image(h,w,c,layer.output);
 }
@@ -15,8 +25,8 @@ image get_convolutional_image(convolutional_layer layer)
 image get_convolutional_delta(convolutional_layer layer)
 {
     int h,w,c;
-    h = layer.out_h;
-    w = layer.out_w;
+    h = convolutional_out_height(layer);
+    w = convolutional_out_width(layer);
     c = layer.n;
     return float_to_image(h,w,c,layer.delta);
 }
@@ -24,7 +34,6 @@ image get_convolutional_delta(convolutional_layer layer)
 convolutional_layer *make_convolutional_layer(int h, int w, int c, int n, int size, int stride, ACTIVATION activation)
 {
     int i;
-    int out_h,out_w;
     size = 2*(size/2)+1; //HA! And you thought you'd use an even sized filter...
     convolutional_layer *layer = calloc(1, sizeof(convolutional_layer));
     layer->h = h;
@@ -47,15 +56,13 @@ convolutional_layer *make_convolutional_layer(int h, int w, int c, int n, int si
         //layer->biases[i] = rand_normal()*scale + scale;
         layer->biases[i] = 0;
     }
-    out_h = (h-size)/stride + 1;
-    out_w = (w-size)/stride + 1;
+    int out_h = (h-size)/stride + 1;
+    int out_w = (w-size)/stride + 1;
 
     layer->col_image = calloc(out_h*out_w*size*size*c, sizeof(float));
     layer->output = calloc(out_h * out_w * n, sizeof(float));
     layer->delta  = calloc(out_h * out_w * n, sizeof(float));
     layer->activation = activation;
-    layer->out_h = out_h;
-    layer->out_w = out_w;
 
     fprintf(stderr, "Convolutional Layer: %d x %d x %d image, %d filters -> %d x %d x %d image\n", h,w,c,n, out_h, out_w, n);
     srand(0);
@@ -90,7 +97,10 @@ void forward_convolutional_layer(const convolutional_layer layer, float *in)
 void gradient_delta_convolutional_layer(convolutional_layer layer)
 {
     int i;
-    for(i = 0; i < layer.out_h*layer.out_w*layer.n; ++i){
+    int size = convolutional_out_height(layer)
+                *convolutional_out_width(layer)
+                *layer.n;
+    for(i = 0; i < size; ++i){
         layer.delta[i] *= gradient(layer.output[i], layer.activation);
     }
 }
@@ -98,7 +108,8 @@ void gradient_delta_convolutional_layer(convolutional_layer layer)
 void learn_bias_convolutional_layer(convolutional_layer layer)
 {
     int i,j;
-    int size = layer.out_h*layer.out_w;
+    int size = convolutional_out_height(layer)
+                *convolutional_out_width(layer);
     for(i = 0; i < layer.n; ++i){
         float sum = 0;
         for(j = 0; j < size; ++j){

src/convolutional_layer.h

@@ -6,7 +6,6 @@
 
 typedef struct {
     int h,w,c;
-    int out_h, out_w, out_c;
     int n;
     int size;
     int stride;

src/data.c

@@ -1,5 +1,4 @@
 #include "data.h"
-#include "list.h"
 #include "utils.h"
 #include "image.h"
 

src/data.h

@@ -2,6 +2,7 @@
 #define DATA_H
 
 #include "matrix.h"
+#include "list.h"
 
 typedef struct{
     matrix X;
@@ -16,6 +17,7 @@ data load_data_image_pathfile_part(char *filename, int part, int total,
                                     char **labels, int k, int h, int w);
 data load_data_image_pathfile_random(char *filename, int n, char **labels, 
                                         int k, int h, int w);
+list *get_paths(char *filename);
 data load_categorical_data_csv(char *filename, int target, int k);
 void normalize_data_rows(data d);
 void scale_data_rows(data d, float s);

src/image.c

@@ -342,21 +342,11 @@ IplImage* resizeImage(const IplImage *origImg, int newHeight, int newWidth,
     return outImg;
 }
 
-image load_image(char *filename, int h, int w)
+image ipl_to_image(IplImage* src)
 {
-    IplImage* src = 0;
-    if( (src = cvLoadImage(filename,-1)) == 0 )
-    {
-        printf("Cannot load file image %s\n", filename);
-        exit(0);
-    }
-    cvShowImage("Orig", src);
-    IplImage *resized = resizeImage(src, h, w, 1);
-    cvShowImage("Sized", resized);
-    cvWaitKey(0);
-    cvReleaseImage(&src);
-    src = resized;
     unsigned char *data = (unsigned char *)src->imageData;
+    int h = src->height;
+    int w = src->width;
     int c = src->nChannels;
     int step = src->widthStep;
     image out = make_image(h,w,c);
@@ -369,6 +359,21 @@ image load_image(char *filename, int h, int w)
             }
         }
     }
+    return out;
+}
+
+image load_image(char *filename, int h, int w)
+{
+    IplImage* src = 0;
+    if( (src = cvLoadImage(filename,-1)) == 0 )
+    {
+        printf("Cannot load file image %s\n", filename);
+        exit(0);
+    }
+    IplImage *resized = resizeImage(src, h, w, 1);
+    cvReleaseImage(&src);
+    src = resized;
+    image out = ipl_to_image(src);
     cvReleaseImage(&src);
     return out;
 }

src/image.h

@@ -34,6 +34,7 @@ image make_random_kernel(int size, int c, float scale);
 image float_to_image(int h, int w, int c, float *data);
 image copy_image(image p);
 image load_image(char *filename, int h, int w);
+image ipl_to_image(IplImage* src);
 
 float get_pixel(image m, int x, int y, int c);
 float get_pixel_extend(image m, int x, int y, int c);

src/network.c

@@ -331,6 +331,34 @@ int get_network_output_size_layer(network net, int i)
     return 0;
 }
 
+int reset_network_size(network net, int h, int w, int c)
+{
+    int i;
+    for (i = 0; i < net.n; ++i){
+        if(net.types[i] == CONVOLUTIONAL){
+            convolutional_layer *layer = (convolutional_layer *)net.layers[i];
+            layer->h = h;
+            layer->w = w;
+            layer->c = c;
+            image output = get_convolutional_image(*layer);
+            h = output.h;
+            w = output.w;
+            c = output.c;
+        }
+        else if(net.types[i] == MAXPOOL){
+            maxpool_layer *layer = (maxpool_layer *)net.layers[i];
+            layer->h = h;
+            layer->w = w;
+            layer->c = c;
+            image output = get_maxpool_image(*layer);
+            h = output.h;
+            w = output.w;
+            c = output.c;
+        }
+    }
+    return 0;
+}
+
 int get_network_output_size(network net)
 {
     int i = net.n-1;

src/network.h

@@ -41,6 +41,7 @@ int get_predicted_class_network(network net);
 void print_network(network net);
 void visualize_network(network net);
 void save_network(network net, char *filename);
+int reset_network_size(network net, int h, int w, int c);
 
 #endif
 

src/tests.c

@@ -366,20 +366,21 @@ void test_im2row()
 
 void train_VOC()
 {
-    network net = parse_network_cfg("cfg/voc_backup_ramp_80.cfg");
+    network net = parse_network_cfg("cfg/voc_backup_sig_20.cfg");
     srand(2222222);
-    int i = 0;
+    int i = 20;
     char *labels[] = {"aeroplane","bicycle","bird","boat","bottle","bus","car","cat","chair","cow","diningtable","dog","horse","motorbike","person","pottedplant","sheep","sofa","train","tvmonitor"};
     float lr = .00001;
     float momentum = .9;
     float decay = 0.01;
     while(i++ < 1000 || 1){
-        visualize_network(net);
-        cvWaitKey(100);
         data train = load_data_image_pathfile_random("images/VOC2012/train_paths.txt", 1000, labels, 20, 300, 400);
+
         image im = float_to_image(300, 400, 3,train.X.vals[0]);
         show_image(im, "input");
+        visualize_network(net);
         cvWaitKey(100);
+
         normalize_data_rows(train);
         clock_t start = clock(), end;
         float loss = train_network_sgd(net, train, 1000, lr, momentum, decay);
@@ -388,13 +389,61 @@ void train_VOC()
         free_data(train);
         if(i%10==0){
             char buff[256];
-            sprintf(buff, "cfg/voc_backup_ramp_%d.cfg", i);
+            sprintf(buff, "cfg/voc_backup_sig_%d.cfg", i);
             save_network(net, buff);
         }
         //lr *= .99;
     }
 }
 
+void features_VOC()
+{
+    int i,j;
+    network net = parse_network_cfg("cfg/voc_features.cfg");
+    char *path_file = "images/VOC2012/all_paths.txt";
+    char *out_dir = "voc_features/";
+    list *paths = get_paths(path_file);
+    node *n = paths->front;
+    while(n){
+        char *path = (char *)n->val;
+        char buff[1024];
+        sprintf(buff, "%s%s.txt",out_dir, path);
+        FILE *fp = fopen(buff, "w");
+        if(fp == 0) file_error(buff);
+
+        IplImage* src = 0;
+        if( (src = cvLoadImage(path,-1)) == 0 )
+        {
+            printf("Cannot load file image %s\n", path);
+            exit(0);
+        }
+
+        for(i = 0; i < 10; ++i){
+            int w = 1024 - 90*i; //PICKED WITH CAREFUL CROSS-VALIDATION!!!!
+            int h = (int)((double)w/src->width * src->height);
+            IplImage *sized = cvCreateImage(cvSize(w,h), src->depth, src->nChannels);
+            cvResize(src, sized, CV_INTER_LINEAR);
+            image im = ipl_to_image(sized);
+            reset_network_size(net, im.h, im.w, im.c);
+            forward_network(net, im.data);
+            free_image(im);
+            image out = get_network_image_layer(net, 5);
+            fprintf(fp, "%d, %d, %d\n",out.c, out.h, out.w);
+            for(j = 0; j < out.c*out.h*out.w; ++j){
+                if(j != 0)fprintf(fp, ",");
+                fprintf(fp, "%g", out.data[j]);
+            }
+            fprintf(fp, "\n");
+            out.c = 1;
+            show_image(out, "output");
+            cvWaitKey(10);
+            cvReleaseImage(&sized);
+        }
+        fclose(fp);
+        n = n->next;
+    }
+}
+
 int main()
 {
     //feenableexcept(FE_DIVBYZERO | FE_INVALID | FE_OVERFLOW);
@@ -406,7 +455,8 @@ int main()
     //test_ensemble();
     //test_nist();
     //test_full();
-    train_VOC();
+    //train_VOC();
+    features_VOC();
     //test_random_preprocess();
     //test_random_classify();
     //test_parser();