mirror/darknet
1
0
Fork 0
mirror of https://github.com/pjreddie/darknet.git synced 2023-08-10 21:13:14 +03:00
Code Issues Projects Releases Wiki Activity

big change to images

Browse Source
This commit is contained in:
Joseph Redmon 2015-04-09 15:18:54 -07:00
parent 1fd10265f8
commit 4f50e29365
12 changed files with 136 additions and 489 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
src/captcha.c
View File

@ -23,7 +23,7 @@ void train_captcha(char *cfgfile, char *weightfile)
while(1){
++i;
time=clock();
data train = load_data_captcha(paths, imgs, plist->size, 10, 60, 200);
data train = load_data_captcha(paths, imgs, plist->size, 10, 200, 60);
translate_data_rows(train, -128);
scale_data_rows(train, 1./128);
printf("Loaded: %lf seconds\n", sec(clock()-time));
@ -56,11 +56,11 @@ void decode_captcha(char *cfgfile, char *weightfile)
printf("Enter filename: ");
fgets(filename, 256, stdin);
strtok(filename, "\n");
image im = load_image_color(filename, 57, 300);
image im = load_image_color(filename, 300, 57);
scale_image(im, 1./255.);
float *X = im.data;
float *predictions = network_predict(net, X);
image out = float_to_image(57, 300, 1, predictions);
image out = float_to_image(300, 57, 1, predictions);
show_image(out, "decoded");
cvWaitKey(0);
free_image(im);
@ -87,7 +87,7 @@ void encode_captcha(char *cfgfile, char *weightfile)
while(1){
++i;
time=clock();
data train = load_data_captcha_encode(paths, imgs, plist->size, 57, 300);
data train = load_data_captcha_encode(paths, imgs, plist->size, 300, 57);
scale_data_rows(train, 1./255);
printf("Loaded: %lf seconds\n", sec(clock()-time));
time=clock();
@ -118,7 +118,7 @@ void validate_captcha(char *cfgfile, char *weightfile)
list *plist = get_paths("/data/captcha/solved.hard");
char **paths = (char **)list_to_array(plist);
int imgs = plist->size;
data valid = load_data_captcha(paths, imgs, 0, 10, 60, 200);
data valid = load_data_captcha(paths, imgs, 0, 10, 200, 60);
translate_data_rows(valid, -128);
scale_data_rows(valid, 1./128);
matrix pred = network_predict_data(net, valid);
@ -157,7 +157,7 @@ void test_captcha(char *cfgfile, char *weightfile)
//printf("Enter filename: ");
fgets(filename, 256, stdin);
strtok(filename, "\n");
image im = load_image_color(filename, 60, 200);
image im = load_image_color(filename, 200, 60);
translate_image(im, -128);
scale_image(im, 1/128.);
float *X = im.data;

36
src/convolutional_layer.c
View File

@ -29,7 +29,7 @@ image get_convolutional_image(convolutional_layer layer)
h = convolutional_out_height(layer);
w = convolutional_out_width(layer);
c = layer.n;
return float_to_image(h,w,c,layer.output);
return float_to_image(w,h,c,layer.output);
}
image get_convolutional_delta(convolutional_layer layer)
@ -38,7 +38,7 @@ image get_convolutional_delta(convolutional_layer layer)
h = convolutional_out_height(layer);
w = convolutional_out_width(layer);
c = layer.n;
return float_to_image(h,w,c,layer.delta);
return float_to_image(w,h,c,layer.delta);
}
convolutional_layer *make_convolutional_layer(int batch, int h, int w, int c, int n, int size, int stride, int pad, ACTIVATION activation)
@ -217,42 +217,22 @@ image get_convolutional_filter(convolutional_layer layer, int i)
int h = layer.size;
int w = layer.size;
int c = layer.c;
return float_to_image(h,w,c,layer.filters+i*h*w*c);
return float_to_image(w,h,c,layer.filters+i*h*w*c);
}
image *weighted_sum_filters(convolutional_layer layer, image *prev_filters)
image *get_filters(convolutional_layer layer)
{
image *filters = calloc(layer.n, sizeof(image));
int i,j,k,c;
if(!prev_filters){
for(i = 0; i < layer.n; ++i){
filters[i] = copy_image(get_convolutional_filter(layer, i));
}
}
else{
image base = prev_filters[0];
for(i = 0; i < layer.n; ++i){
image filter = get_convolutional_filter(layer, i);
filters[i] = make_image(base.h, base.w, base.c);
for(j = 0; j < layer.size; ++j){
for(k = 0; k < layer.size; ++k){
for(c = 0; c < layer.c; ++c){
float weight = get_pixel(filter, j, k, c);
image prev_filter = copy_image(prev_filters[c]);
scale_image(prev_filter, weight);
add_into_image(prev_filter, filters[i], 0,0);
free_image(prev_filter);
}
}
}
}
int i;
for(i = 0; i < layer.n; ++i){
filters[i] = copy_image(get_convolutional_filter(layer, i));
}
return filters;
}
image *visualize_convolutional_layer(convolutional_layer layer, char *window, image *prev_filters)
{
image *single_filters = weighted_sum_filters(layer, 0);
image *single_filters = get_filters(layer);
show_images(single_filters, layer.n, window);
image delta = get_convolutional_image(layer);

2
src/crop_layer.c
View File

@ -7,7 +7,7 @@ image get_crop_image(crop_layer layer)
int h = layer.crop_height;
int w = layer.crop_width;
int c = layer.c;
return float_to_image(h,w,c,layer.output);
return float_to_image(w,h,c,layer.output);
}
crop_layer *make_crop_layer(int batch, int h, int w, int c, int crop_height, int crop_width, int flip)

45
src/data.c
View File

@ -47,7 +47,7 @@ char **get_random_paths(char **paths, int n, int m)
return random_paths;
}
matrix load_image_paths(char **paths, int n, int h, int w)
matrix load_image_paths(char **paths, int n, int w, int h)
{
int i;
matrix X;
@ -56,7 +56,7 @@ matrix load_image_paths(char **paths, int n, int h, int w)
X.cols = 0;
for(i = 0; i < n; ++i){
image im = load_image_color(paths[i], h, w);
image im = load_image_color(paths[i], w, h);
X.vals[i] = im.data;
X.cols = im.h*im.w*im.c;
}
@ -207,12 +207,12 @@ void fill_truth_captcha(char *path, int n, float *truth)
}
}
data load_data_captcha(char **paths, int n, int m, int k, int h, int w)
data load_data_captcha(char **paths, int n, int m, int k, int w, int h)
{
if(m) paths = get_random_paths(paths, n, m);
data d;
d.shallow = 0;
d.X = load_image_paths(paths, n, h, w);
d.X = load_image_paths(paths, n, w, h);
d.y = make_matrix(n, k*NUMCHARS);
int i;
for(i = 0; i < n; ++i){
@ -222,12 +222,12 @@ data load_data_captcha(char **paths, int n, int m, int k, int h, int w)
return d;
}
data load_data_captcha_encode(char **paths, int n, int m, int h, int w)
data load_data_captcha_encode(char **paths, int n, int m, int w, int h)
{
if(m) paths = get_random_paths(paths, n, m);
data d;
d.shallow = 0;
d.X = load_image_paths(paths, n, h, w);
d.X = load_image_paths(paths, n, w, h);
d.X.cols = 17100;
d.y = d.X;
if(m) free(paths);
@ -258,21 +258,6 @@ matrix load_labels_paths(char **paths, int n, char **labels, int k)
return y;
}
data load_data_image_pathfile(char *filename, char **labels, int k, int h, int w)
{
list *plist = get_paths(filename);
char **paths = (char **)list_to_array(plist);
int n = plist->size;
data d;
d.shallow = 0;
d.X = load_image_paths(paths, n, h, w);
d.y = load_labels_paths(paths, n, labels, k);
free_list_contents(plist);
free_list(plist);
free(paths);
return d;
}
char **get_labels(char *filename)
{
list *plist = get_paths(filename);
@ -292,7 +277,7 @@ void free_data(data d)
}
}
data load_data_detection_jitter_random(int n, char **paths, int m, int classes, int h, int w, int num_boxes, int background)
data load_data_detection_jitter_random(int n, char **paths, int m, int classes, int w, int h, int num_boxes, int background)
{
char **random_paths = get_random_paths(paths, n, m);
int i;
@ -325,12 +310,12 @@ data load_data_detection_jitter_random(int n, char **paths, int m, int classes,
float sy = (float)sheight / oh;
int flip = rand()%2;
image cropped = crop_image(orig, ptop, pleft, sheight, swidth);
image cropped = crop_image(orig, pleft, ptop, swidth, sheight);
float dx = ((float)pleft/ow)/sx;
float dy = ((float)ptop /oh)/sy;
free_image(orig);
image sized = resize_image(cropped, h, w);
image sized = resize_image(cropped, w, h);
free_image(cropped);
if(flip) flip_image(sized);
d.X.vals[i] = sized.data;
@ -345,14 +330,14 @@ void *load_detection_thread(void *ptr)
{
printf("Loading data: %d\n", rand());
struct load_args a = *(struct load_args*)ptr;
*a.d = load_data_detection_jitter_random(a.n, a.paths, a.m, a.classes, a.h, a.w, a.num_boxes, a.background);
*a.d = load_data_detection_jitter_random(a.n, a.paths, a.m, a.classes, a.w, a.h, a.num_boxes, a.background);
translate_data_rows(*a.d, -128);
scale_data_rows(*a.d, 1./128);
free(ptr);
return 0;
}
pthread_t load_data_detection_thread(int n, char **paths, int m, int classes, int h, int w, int nh, int nw, int background, data *d)
pthread_t load_data_detection_thread(int n, char **paths, int m, int classes, int w, int h, int nh, int nw, int background, data *d)
{
pthread_t thread;
struct load_args *args = calloc(1, sizeof(struct load_args));
@ -373,12 +358,12 @@ pthread_t load_data_detection_thread(int n, char **paths, int m, int classes, in
return thread;
}
data load_data(char **paths, int n, int m, char **labels, int k, int h, int w)
data load_data(char **paths, int n, int m, char **labels, int k, int w, int h)
{
if(m) paths = get_random_paths(paths, n, m);
data d;
d.shallow = 0;
d.X = load_image_paths(paths, n, h, w);
d.X = load_image_paths(paths, n, w, h);
d.y = load_labels_paths(paths, n, labels, k);
if(m) free(paths);
return d;
@ -387,14 +372,14 @@ data load_data(char **paths, int n, int m, char **labels, int k, int h, int w)
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);
*a.d = load_data(a.paths, a.n, a.m, a.labels, a.k, a.w, a.h);
translate_data_rows(*a.d, -128);
scale_data_rows(*a.d, 1./128);
free(ptr);
return 0;
}
pthread_t load_data_thread(char **paths, int n, int m, char **labels, int k, int h, int w, data *d)
pthread_t load_data_thread(char **paths, int n, int m, char **labels, int k, int w, int h, data *d)
{
pthread_t thread;
struct load_args *args = calloc(1, sizeof(struct load_args));

14
src/data.h
View File

@ -27,17 +27,17 @@ typedef struct{
void free_data(data d);
void print_letters(float *pred, int n);
data load_data_captcha(char **paths, int n, int m, int k, int h, int w);
data load_data_captcha_encode(char **paths, int n, int m, int h, int w);
data load_data(char **paths, int n, int m, char **labels, int k, int h, int w);
pthread_t load_data_thread(char **paths, int n, int m, char **labels, int k, int h, int w, data *d);
data load_data_captcha(char **paths, int n, int m, int k, int w, int h);
data load_data_captcha_encode(char **paths, int n, int m, int w, int h);
data load_data(char **paths, int n, int m, char **labels, int k, int w, int h);
pthread_t load_data_thread(char **paths, int n, int m, char **labels, int k, int w, int h, data *d);
pthread_t load_data_detection_thread(int n, char **paths, int m, int classes, int h, int w, int nh, int nw, int background, data *d);
data load_data_detection_jitter_random(int n, char **paths, int m, int classes, int h, int w, int num_boxes, int background);
pthread_t load_data_detection_thread(int n, char **paths, int m, int classes, int w, int h, int nh, int nw, int background, data *d);
data load_data_detection_jitter_random(int n, char **paths, int m, int classes, int w, int h, int num_boxes, int background);
data load_data_image_pathfile(char *filename, char **labels, int k, int h, int w);
data load_cifar10_data(char *filename);
data load_all_cifar10();
list *get_paths(char *filename);
char **get_labels(char *filename);
void get_random_batch(data d, int n, float *X, float *y);

4
src/deconvolutional_layer.c
View File

@ -31,7 +31,7 @@ image get_deconvolutional_image(deconvolutional_layer layer)
h = deconvolutional_out_height(layer);
w = deconvolutional_out_width(layer);
c = layer.n;
return float_to_image(h,w,c,layer.output);
return float_to_image(w,h,c,layer.output);
}
image get_deconvolutional_delta(deconvolutional_layer layer)
@ -40,7 +40,7 @@ image get_deconvolutional_delta(deconvolutional_layer layer)
h = deconvolutional_out_height(layer);
w = deconvolutional_out_width(layer);
c = layer.n;
return float_to_image(h,w,c,layer.delta);
return float_to_image(w,h,c,layer.delta);
}
deconvolutional_layer *make_deconvolutional_layer(int batch, int h, int w, int c, int n, int size, int stride, ACTIVATION activation)

13
src/detection.c
View File

@ -83,14 +83,14 @@ void train_detection(char *cfgfile, char *weightfile)
plist = get_paths("/home/pjreddie/data/voc/trainall.txt");
}
paths = (char **)list_to_array(plist);
pthread_t load_thread = load_data_detection_thread(imgs, paths, plist->size, classes, net.h, net.w, side, side, background, &buffer);
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(1){
i += 1;
time=clock();
pthread_join(load_thread, 0);
train = buffer;
load_thread = load_data_detection_thread(imgs, paths, plist->size, classes, net.h, net.w, side, side, background, &buffer);
load_thread = load_data_detection_thread(imgs, paths, plist->size, classes, net.w, net.h, side, side, background, &buffer);
/*
image im = float_to_image(im_dim, im_dim, 3, train.X.vals[114]);
@ -124,6 +124,7 @@ void validate_detection(char *cfgfile, char *weightfile)
srand(time(0));
list *plist = get_paths("/home/pjreddie/data/voc/val.txt");
//list *plist = get_paths("/home/pjreddie/data/voc/val.expanded.txt");
//list *plist = get_paths("/home/pjreddie/data/voc/train.txt");
char **paths = (char **)list_to_array(plist);
@ -142,7 +143,7 @@ void validate_detection(char *cfgfile, char *weightfile)
fprintf(stderr, "%d\n", m);
data val, buffer;
pthread_t load_thread = load_data_thread(paths, num, 0, 0, num_output, net.h, net.w, &buffer);
pthread_t load_thread = load_data_thread(paths, num, 0, 0, num_output, net.w, net.h, &buffer);
clock_t time;
for(i = 1; i <= splits; ++i){
time=clock();
@ -151,7 +152,7 @@ void validate_detection(char *cfgfile, char *weightfile)
num = (i+1)*m/splits - i*m/splits;
char **part = paths+(i*m/splits);
if(i != splits) load_thread = load_data_thread(part, num, 0, 0, num_output, net.h, net.w, &buffer);
if(i != splits) load_thread = load_data_thread(part, num, 0, 0, num_output, net.w, net.h, &buffer);
fprintf(stderr, "%d: Loaded: %lf seconds\n", i, sec(clock()-time));
matrix pred = network_predict_data(net, val);
@ -171,7 +172,9 @@ void validate_detection(char *cfgfile, char *weightfile)
h = h*h;
float w = pred.vals[j][ci + 3]; //* distance_from_edge(col, num_boxes);
w = w*w;
printf("%d %d %f %f %f %f %f\n", (i-1)*m/splits + j, class, scale*pred.vals[j][k+class+background+nuisance], y, x, h, w);
float prob = scale*pred.vals[j][k+class+background+nuisance];
if(prob < .001) continue;
printf("%d %d %f %f %f %f %f\n", (i-1)*m/splits + j, class, prob, y, x, h, w);
}
}
}

451
src/image.c
View File

@ -53,21 +53,6 @@ void draw_box(image a, int x1, int y1, int x2, int y2, float r, float g, float b
}
}
void jitter_image(image a, int h, int w, int dh, int dw)
{
int i,j,k;
for(k = 0; k < a.c; ++k){
for(i = 0; i < h; ++i){
for(j = 0; j < w; ++j){
int src = j + dw + (i+dh)*a.w + k*a.w*a.h;
int dst = j + i*w + k*w*h;
//printf("%d %d\n", src, dst);
a.data[dst] = a.data[src];
}
}
}
}
void flip_image(image a)
{
int i,j,k;
@ -87,7 +72,7 @@ void flip_image(image a)
image image_distance(image a, image b)
{
int i,j;
image dist = make_image(a.h, a.w, 1);
image dist = make_image(a.w, a.h, 1);
for(i = 0; i < a.c; ++i){
for(j = 0; j < a.h*a.w; ++j){
dist.data[j] += pow(a.data[i*a.h*a.w+j]-b.data[i*a.h*a.w+j],2);
@ -99,20 +84,14 @@ image image_distance(image a, image b)
return dist;
}
void subtract_image(image a, image b)
void embed_image(image source, image dest, int dx, int dy)
{
int i;
for(i = 0; i < a.h*a.w*a.c; ++i) a.data[i] -= b.data[i];
}
void embed_image(image source, image dest, int h, int w)
{
int i,j,k;
int x,y,k;
for(k = 0; k < source.c; ++k){
for(i = 0; i < source.h; ++i){
for(j = 0; j < source.w; ++j){
float val = get_pixel(source, i,j,k);
set_pixel(dest, h+i, w+j, k, val);
for(y = 0; y < source.h; ++y){
for(x = 0; x < source.w; ++x){
float val = get_pixel(source, x,y,k);
set_pixel(dest, dx+x, dy+y, k, val);
}
}
}
@ -122,22 +101,17 @@ image collapse_image_layers(image source, int border)
{
int h = source.h;
h = (h+border)*source.c - border;
image dest = make_image(h, source.w, 1);
image dest = make_image(source.w, h, 1);
int i;
for(i = 0; i < source.c; ++i){
image layer = get_image_layer(source, i);
int h_offset = i*(source.h+border);
embed_image(layer, dest, h_offset, 0);
embed_image(layer, dest, 0, h_offset);
free_image(layer);
}
return dest;
}
void z_normalize_image(image p)
{
normalize_array(p.data, p.h*p.w*p.c);
}
void normalize_image(image p)
{
float *min = calloc(p.c, sizeof(float));
@ -167,24 +141,6 @@ void normalize_image(image p)
free(max);
}
float avg_image_layer(image m, int l)
{
int i;
float sum = 0;
for(i = 0; i < m.h*m.w; ++i){
sum += m.data[l*m.h*m.w + i];
}
return sum/(m.h*m.w);
}
void threshold_image(image p, float t)
{
int i;
for(i = 0; i < p.w*p.h*p.c; ++i){
if(p.data[i] < t) p.data[i] = 0;
}
}
image copy_image(image p)
{
image copy = p;
@ -196,7 +152,7 @@ image copy_image(image p)
void show_image(image p, char *name)
{
int i,j,k;
int x,y,k;
image copy = copy_image(p);
normalize_image(copy);
@ -209,10 +165,10 @@ void show_image(image p, char *name)
cvNamedWindow(buff, CV_WINDOW_AUTOSIZE);
//cvMoveWindow(buff, 100*(windows%10) + 200*(windows/10), 100*(windows%10));
++windows;
for(i = 0; i < p.h; ++i){
for(j = 0; j < p.w; ++j){
for(y = 0; y < p.h; ++y){
for(x = 0; x < p.w; ++x){
for(k= 0; k < p.c; ++k){
disp->imageData[i*step + j*p.c + k] = (unsigned char)(get_pixel(copy,i,j,k)*255);
disp->imageData[y*step + x*p.c + k] = (unsigned char)(get_pixel(copy,x,y,k)*255);
}
}
}
@ -235,7 +191,7 @@ void show_image(image p, char *name)
void save_image(image p, char *name)
{
int i,j,k;
int x,y,k;
image copy = copy_image(p);
normalize_image(copy);
@ -245,10 +201,10 @@ void save_image(image p, char *name)
IplImage *disp = cvCreateImage(cvSize(p.w,p.h), IPL_DEPTH_8U, p.c);
int step = disp->widthStep;
for(i = 0; i < p.h; ++i){
for(j = 0; j < p.w; ++j){
for(y = 0; y < p.h; ++y){
for(x = 0; x < p.w; ++x){
for(k= 0; k < p.c; ++k){
disp->imageData[i*step + j*p.c + k] = (unsigned char)(get_pixel(copy,i,j,k)*255);
disp->imageData[y*step + x*p.c + k] = (unsigned char)(get_pixel(copy,x,y,k)*255);
}
}
}
@ -276,7 +232,7 @@ void show_image_collapsed(image p, char *name)
free_image(c);
}
image make_empty_image(int h, int w, int c)
image make_empty_image(int w, int h, int c)
{
image out;
out.data = 0;
@ -286,30 +242,20 @@ image make_empty_image(int h, int w, int c)
return out;
}
image make_image(int h, int w, int c)
image make_image(int w, int h, int c)
{
image out = make_empty_image(h,w,c);
image out = make_empty_image(w,h,c);
out.data = calloc(h*w*c, sizeof(float));
return out;
}
image float_to_image(int h, int w, int c, float *data)
image float_to_image(int w, int h, int c, float *data)
{
image out = make_empty_image(h,w,c);
image out = make_empty_image(w,h,c);
out.data = data;
return out;
}
void zero_image(image m)
{
memset(m.data, 0, m.h*m.w*m.c*sizeof(float));
}
void zero_channel(image m, int c)
{
memset(&(m.data[c*m.h*m.w]), 0, m.h*m.w*sizeof(float));
}
void rotate_image(image m)
{
int i,j;
@ -322,29 +268,6 @@ void rotate_image(image m)
}
}
image make_random_image(int h, int w, int c)
{
image out = make_image(h,w,c);
int i;
for(i = 0; i < h*w*c; ++i){
out.data[i] = rand_normal();
//out.data[i] = rand()%3;
}
return out;
}
void add_into_image(image src, image dest, int h, int w)
{
int i,j,k;
for(k = 0; k < src.c; ++k){
for(i = 0; i < src.h; ++i){
for(j = 0; j < src.w; ++j){
add_pixel(dest, h+i, w+j, k, get_pixel(src, i, j, k));
}
}
}
}
void translate_image(image m, float s)
{
int i;
@ -357,24 +280,6 @@ void scale_image(image m, float s)
for(i = 0; i < m.h*m.w*m.c; ++i) m.data[i] *= s;
}
image make_random_kernel(int size, int c, float scale)
{
int pad;
if((pad=(size%2==0))) ++size;
image out = make_random_image(size,size,c);
scale_image(out, scale);
int i,k;
if(pad){
for(k = 0; k < out.c; ++k){
for(i = 0; i < size; ++i) {
set_pixel(out, i, 0, k, 0);
set_pixel(out, 0, i, k, 0);
}
}
}
return out;
}
image ipl_to_image(IplImage* src)
{
unsigned char *data = (unsigned char *)src->imageData;
@ -382,7 +287,7 @@ image ipl_to_image(IplImage* src)
int w = src->width;
int c = src->nChannels;
int step = src->widthStep;
image out = make_image(h,w,c);
image out = make_image(w, h, c);
int i, j, k, count=0;;
for(k= 0; k < c; ++k){
@ -395,47 +300,55 @@ image ipl_to_image(IplImage* src)
return out;
}
image crop_image(image im, int dr, int dc, int h, int w)
image crop_image(image im, int dx, int dy, int w, int h)
{
image cropped = make_image(h, w, im.c);
image cropped = make_image(w, h, im.c);
int i, j, k;
for(k = 0; k < im.c; ++k){
for(j = 0; j < h; ++j){
for(i = 0; i < w; ++i){
int r = j + dr;
int c = i + dc;
int r = j + dy;
int c = i + dx;
float val = 128;
if (r >= 0 && r < im.h && c >= 0 && c < im.w) {
val = get_pixel(im, r, c, k);
val = get_pixel(im, c, r, k);
}
set_pixel(cropped, j, i, k, val);
set_pixel(cropped, i, j, k, val);
}
}
}
return cropped;
}
// #wikipedia
image resize_image(image im, int h, int w)
float billinear_interpolate(image im, float x, float y, int c)
{
image resized = make_image(h, w, im.c);
int ix = (int) x;
int iy = (int) y;
float dx = x - ix;
float dy = y - iy;
float val = (1-dy) * (1-dx) * get_pixel_extend(im, ix, iy, c) +
dy * (1-dx) * get_pixel_extend(im, ix, iy+1, c) +
(1-dy) * dx * get_pixel_extend(im, ix+1, iy, c) +
dy * dx * get_pixel_extend(im, ix+1, iy+1, c);
return val;
}
// #wikipedia
image resize_image(image im, int w, int h)
{
image resized = make_image(w, h, im.c);
int r, c, k;
float h_scale = (float)(im.h - 1) / (h - 1) - .00001;
float w_scale = (float)(im.w - 1) / (w - 1) - .00001;
float w_scale = (float)(im.w - 1) / (w - 1);
float h_scale = (float)(im.h - 1) / (h - 1);
for(k = 0; k < im.c; ++k){
for(r = 0; r < h; ++r){
for(c = 0; c < w; ++c){
float sr = r*h_scale;
float sc = c*w_scale;
int ir = (int)sr;
int ic = (int)sc;
float x = sr-ir;
float y = sc-ic;
float val = (1-x) * (1-y) * get_pixel(im, ir, ic, k) +
x * (1-y) * get_pixel(im, ir+1, ic, k) +
(1-x) * y * get_pixel(im, ir, ic+1, k) +
x * y * get_pixel(im, ir+1, ic+1, k);
set_pixel(resized, r, c, k, val);
float sx = c*w_scale;
float sy = r*h_scale;
float val = billinear_interpolate(im, sx, sy, k);
set_pixel(resized, c, r, k, val);
}
}
}
@ -445,10 +358,10 @@ image resize_image(image im, int h, int w)
void test_resize(char *filename)
{
image im = load_image(filename, 0,0);
image small = resize_image(im, 63, 65);
image big = resize_image(im, 512, 513);
image crop = crop_image(im, 10, 50, 100, 100);
image crop2 = crop_image(im, -50, -30, 400, 291);
image small = resize_image(im, 65, 63);
image big = resize_image(im, 513, 512);
image crop = crop_image(im, 50, 10, 100, 100);
image crop2 = crop_image(im, -30, -50, 291, 400);
show_image(im, "original");
show_image(small, "smaller");
show_image(big, "bigger");
@ -457,7 +370,7 @@ void test_resize(char *filename)
cvWaitKey(0);
}
image load_image_color(char *filename, int h, int w)
image load_image_color(char *filename, int w, int h)
{
IplImage* src = 0;
if( (src = cvLoadImage(filename, 1)) == 0 )
@ -467,7 +380,7 @@ image load_image_color(char *filename, int h, int w)
}
image out = ipl_to_image(src);
if((h && w) && (h != out.h || w != out.w)){
image resized = resize_image(out, h, w);
image resized = resize_image(out, w, h);
free_image(out);
out = resized;
}
@ -475,7 +388,7 @@ image load_image_color(char *filename, int h, int w)
return out;
}
image load_image(char *filename, int h, int w)
image load_image(char *filename, int w, int h)
{
IplImage* src = 0;
if( (src = cvLoadImage(filename,-1)) == 0 )
@ -485,7 +398,7 @@ image load_image(char *filename, int h, int w)
}
image out = ipl_to_image(src);
if((h && w) && (h != out.h || w != out.w)){
image resized = resize_image(out, h, w);
image resized = resize_image(out, w, h);
free_image(out);
out = resized;
}
@ -495,209 +408,28 @@ image load_image(char *filename, int h, int w)
image get_image_layer(image m, int l)
{
image out = make_image(m.h, m.w, 1);
image out = make_image(m.w, m.h, 1);
int i;
for(i = 0; i < m.h*m.w; ++i){
out.data[i] = m.data[i+l*m.h*m.w];
}
return out;
}
image get_sub_image(image m, int h, int w, int dh, int dw)
{
image out = make_image(dh, dw, m.c);
int i,j,k;
for(k = 0; k < out.c; ++k){
for(i = 0; i < dh; ++i){
for(j = 0; j < dw; ++j){
float val = get_pixel(m, h+i, w+j, k);
set_pixel(out, i, j, k, val);
}
}
}
return out;
}
float get_pixel(image m, int x, int y, int c)
{
assert(x < m.h && y < m.w && c < m.c);
return m.data[c*m.h*m.w + x*m.w + y];
assert(x < m.w && y < m.h && c < m.c);
return m.data[c*m.h*m.w + y*m.w + x];
}
float get_pixel_extend(image m, int x, int y, int c)
{
if(x < 0 || x >= m.h || y < 0 || y >= m.w || c < 0 || c >= m.c) return 0;
if(x < 0 || x >= m.w || y < 0 || y >= m.h || c < 0 || c >= m.c) return 0;
return get_pixel(m, x, y, c);
}
void set_pixel(image m, int x, int y, int c, float val)
{
assert(x < m.h && y < m.w && c < m.c);
m.data[c*m.h*m.w + x*m.w + y] = val;
}
void set_pixel_extend(image m, int x, int y, int c, float val)
{
if(x < 0 || x >= m.h || y < 0 || y >= m.w || c < 0 || c >= m.c) return;
set_pixel(m, x, y, c, val);
}
void add_pixel(image m, int x, int y, int c, float val)
{
assert(x < m.h && y < m.w && c < m.c);
m.data[c*m.h*m.w + x*m.w + y] += val;
}
void add_pixel_extend(image m, int x, int y, int c, float val)
{
if(x < 0 || x >= m.h || y < 0 || y >= m.w || c < 0 || c >= m.c) return;
add_pixel(m, x, y, c, val);
}
void two_d_convolve(image m, int mc, image kernel, int kc, int stride, image out, int oc, int edge)
{
int x,y,i,j;
int xstart, xend, ystart, yend;
if(edge){
xstart = ystart = 0;
xend = m.h;
yend = m.w;
}else{
xstart = kernel.h/2;
ystart = kernel.w/2;
xend = m.h-kernel.h/2;
yend = m.w - kernel.w/2;
}
for(x = xstart; x < xend; x += stride){
for(y = ystart; y < yend; y += stride){
float sum = 0;
for(i = 0; i < kernel.h; ++i){
for(j = 0; j < kernel.w; ++j){
sum += get_pixel(kernel, i, j, kc)*get_pixel_extend(m, x+i-kernel.h/2, y+j-kernel.w/2, mc);
}
}
add_pixel(out, (x-xstart)/stride, (y-ystart)/stride, oc, sum);
}
}
}
float single_convolve(image m, image kernel, int x, int y)
{
float sum = 0;
int i, j, k;
for(i = 0; i < kernel.h; ++i){
for(j = 0; j < kernel.w; ++j){
for(k = 0; k < kernel.c; ++k){
sum += get_pixel(kernel, i, j, k)*get_pixel_extend(m, x+i-kernel.h/2, y+j-kernel.w/2, k);
}
}
}
return sum;
}
void convolve(image m, image kernel, int stride, int channel, image out, int edge)
{
assert(m.c == kernel.c);
int i;
zero_channel(out, channel);
for(i = 0; i < m.c; ++i){
two_d_convolve(m, i, kernel, i, stride, out, channel, edge);
}
/*
int j;
for(i = 0; i < m.h; i += stride){
for(j = 0; j < m.w; j += stride){
float val = single_convolve(m, kernel, i, j);
set_pixel(out, i/stride, j/stride, channel, val);
}
}
*/
}
void upsample_image(image m, int stride, image out)
{
int i,j,k;
zero_image(out);
for(k = 0; k < m.c; ++k){
for(i = 0; i < m.h; ++i){
for(j = 0; j< m.w; ++j){
float val = get_pixel(m, i, j, k);
set_pixel(out, i*stride, j*stride, k, val);
}
}
}
}
void single_update(image m, image update, int x, int y, float error)
{
int i, j, k;
for(i = 0; i < update.h; ++i){
for(j = 0; j < update.w; ++j){
for(k = 0; k < update.c; ++k){
float val = get_pixel_extend(m, x+i-update.h/2, y+j-update.w/2, k);
add_pixel(update, i, j, k, val*error);
}
}
}
}
void kernel_update(image m, image update, int stride, int channel, image out, int edge)
{
assert(m.c == update.c);
zero_image(update);
int i, j, istart, jstart, iend, jend;
if(edge){
istart = jstart = 0;
iend = m.h;
jend = m.w;
}else{
istart = update.h/2;
jstart = update.w/2;
iend = m.h-update.h/2;
jend = m.w - update.w/2;
}
for(i = istart; i < iend; i += stride){
for(j = jstart; j < jend; j += stride){
float error = get_pixel(out, (i-istart)/stride, (j-jstart)/stride, channel);
single_update(m, update, i, j, error);
}
}
/*
for(i = 0; i < update.h*update.w*update.c; ++i){
update.data[i] /= (m.h/stride)*(m.w/stride);
}
*/
}
void single_back_convolve(image m, image kernel, int x, int y, float val)
{
int i, j, k;
for(i = 0; i < kernel.h; ++i){
for(j = 0; j < kernel.w; ++j){
for(k = 0; k < kernel.c; ++k){
float pval = get_pixel(kernel, i, j, k) * val;
add_pixel_extend(m, x+i-kernel.h/2, y+j-kernel.w/2, k, pval);
}
}
}
}
void back_convolve(image m, image kernel, int stride, int channel, image out, int edge)
{
assert(m.c == kernel.c);
int i, j, istart, jstart, iend, jend;
if(edge){
istart = jstart = 0;
iend = m.h;
jend = m.w;
}else{
istart = kernel.h/2;
jstart = kernel.w/2;
iend = m.h-kernel.h/2;
jend = m.w - kernel.w/2;
}
for(i = istart; i < iend; i += stride){
for(j = jstart; j < jend; j += stride){
float val = get_pixel(out, (i-istart)/stride, (j-jstart)/stride, channel);
single_back_convolve(m, kernel, i, j, val);
}
}
assert(x < m.w && y < m.h && c < m.c);
m.data[c*m.h*m.w + y*m.w + x] = val;
}
void print_image(image m)
@ -730,20 +462,20 @@ image collapse_images_vert(image *ims, int n)
c = 1;
}
image filters = make_image(h,w,c);
image filters = make_image(w, h, c);
int i,j;
for(i = 0; i < n; ++i){
int h_offset = i*(ims[0].h+border);
image copy = copy_image(ims[i]);
//normalize_image(copy);
if(c == 3 && color){
embed_image(copy, filters, h_offset, 0);
embed_image(copy, filters, 0, h_offset);
}
else{
for(j = 0; j < copy.c; ++j){
int w_offset = j*(ims[0].w+border);
image layer = get_image_layer(copy, j);
embed_image(layer, filters, h_offset, w_offset);
embed_image(layer, filters, w_offset, h_offset);
free_image(layer);
}
}
@ -766,20 +498,20 @@ image collapse_images_horz(image *ims, int n)
c = 1;
}
image filters = make_image(h,w,c);
image filters = make_image(w, h, c);
int i,j;
for(i = 0; i < n; ++i){
int w_offset = i*(size+border);
image copy = copy_image(ims[i]);
//normalize_image(copy);
if(c == 3 && color){
embed_image(copy, filters, 0, w_offset);
embed_image(copy, filters, w_offset, 0);
}
else{
for(j = 0; j < copy.c; ++j){
int h_offset = j*(size+border);
image layer = get_image_layer(copy, j);
embed_image(layer, filters, h_offset, w_offset);
embed_image(layer, filters, w_offset, h_offset);
free_image(layer);
}
}
@ -796,43 +528,6 @@ void show_images(image *ims, int n, char *window)
free_image(m);
}
image grid_images(image **ims, int h, int w)
{
int i;
image *rows = calloc(h, sizeof(image));
for(i = 0; i < h; ++i){
rows[i] = collapse_images_horz(ims[i], w);
}
image out = collapse_images_vert(rows, h);
for(i = 0; i < h; ++i){
free_image(rows[i]);
}
free(rows);
return out;
}
void test_grid()
{
int i,j;
int num = 3;
int topk = 3;
image **vizs = calloc(num, sizeof(image*));
for(i = 0; i < num; ++i){
vizs[i] = calloc(topk, sizeof(image));
for(j = 0; j < topk; ++j) vizs[i][j] = make_image(3,3,3);
}
image grid = grid_images(vizs, num, topk);
save_image(grid, "Test Grid");
free_image(grid);
}
void show_images_grid(image **ims, int h, int w, char *window)
{
image out = grid_images(ims, h, w);
show_image(out, window);
free_image(out);
}
void free_image(image m)
{
free(m.data);

39
src/image.h
View File

@ -12,61 +12,44 @@ typedef struct {
} image;
float get_color(int c, int x, int max);
void jitter_image(image a, int h, int w, int dh, int dw);
void flip_image(image a);
void draw_box(image a, int x1, int y1, int x2, int y2, float r, float g, float b);
image image_distance(image a, image b);
void scale_image(image m, float s);
image crop_image(image im, int dr, int dc, int h, int w);
image resize_image(image im, int h, int w);
image crop_image(image im, int dx, int dy, int w, int h);
image resize_image(image im, int w, int h);
void translate_image(image m, float s);
void normalize_image(image p);
void z_normalize_image(image p);
void threshold_image(image p, float t);
void zero_image(image m);
void rotate_image(image m);
void subtract_image(image a, image b);
float avg_image_layer(image m, int l);
void embed_image(image source, image dest, int h, int w);
void add_into_image(image src, image dest, int h, int w);
void embed_image(image source, image dest, int dx, int dy);
image collapse_image_layers(image source, int border);
image collapse_images_horz(image *ims, int n);
image collapse_images_vert(image *ims, int n);
image get_sub_image(image m, int h, int w, int dh, int dw);
void show_image(image p, char *name);
void save_image(image p, char *name);
void show_images(image *ims, int n, char *window);
void show_image_layers(image p, char *name);
void show_image_collapsed(image p, char *name);
void show_images_grid(image **ims, int h, int w, char *window);
void test_grid();
image grid_images(image **ims, int h, int w);
void print_image(image m);
image make_image(int h, int w, int c);
image make_empty_image(int h, int w, int c);
image make_random_image(int h, int w, int c);
image make_random_kernel(int size, int c, float scale);
image float_to_image(int h, int w, int c, float *data);
image make_image(int w, int h, int c);
image make_empty_image(int w, int h, int c);
image float_to_image(int w, int h, int c, float *data);
image copy_image(image p);
image load_image(char *filename, int h, int w);
image load_image_color(char *filename, int h, int w);
image load_image(char *filename, int w, int h);
image load_image_color(char *filename, int w, int h);
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);
void add_pixel(image m, int x, int y, int c, float val);
void set_pixel(image m, int x, int y, int c, float val);
image get_image_layer(image m, int l);
void two_d_convolve(image m, int mc, image kernel, int kc, int stride, image out, int oc, int edge);
void upsample_image(image m, int stride, image out);
void convolve(image m, image kernel, int stride, int channel, image out, int edge);
void back_convolve(image m, image kernel, int stride, int channel, image out, int edge);
void kernel_update(image m, image update, int stride, int channel, image out, int edge);
void free_image(image m);
void test_resize(char *filename);
#endif

4
src/maxpool_layer.c
View File

@ -7,7 +7,7 @@ image get_maxpool_image(maxpool_layer layer)
int h = (layer.h-1)/layer.stride + 1;
int w = (layer.w-1)/layer.stride + 1;
int c = layer.c;
return float_to_image(h,w,c,layer.output);
return float_to_image(w,h,c,layer.output);
}
image get_maxpool_delta(maxpool_layer layer)
@ -15,7 +15,7 @@ image get_maxpool_delta(maxpool_layer layer)
int h = (layer.h-1)/layer.stride + 1;
int w = (layer.w-1)/layer.stride + 1;
int c = layer.c;
return float_to_image(h,w,c,layer.delta);
return float_to_image(w,h,c,layer.delta);
}
maxpool_layer *make_maxpool_layer(int batch, int h, int w, int c, int size, int stride)

4
src/normalization_layer.c
View File

@ -6,7 +6,7 @@ image get_normalization_image(normalization_layer layer)
int h = layer.h;
int w = layer.w;
int c = layer.c;
return float_to_image(h,w,c,layer.output);
return float_to_image(w,h,c,layer.output);
}
image get_normalization_delta(normalization_layer layer)
@ -14,7 +14,7 @@ image get_normalization_delta(normalization_layer layer)
int h = layer.h;
int w = layer.w;
int c = layer.c;
return float_to_image(h,w,c,layer.delta);
return float_to_image(w,h,c,layer.delta);
}
normalization_layer *make_normalization_layer(int batch, int h, int w, int c, int size, float alpha, float beta, float kappa)

1
src/parser.h
View File

@ -6,5 +6,6 @@ network parse_network_cfg(char *filename);
void save_network(network net, char *filename);
void save_weights(network net, char *filename);
void load_weights(network *net, char *filename);
void load_weights_upto(network *net, char *filename, int cutoff);
#endif