From d97331b88ff3d50035b1e22c9d0eb671b61227e3 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Wed, 15 Apr 2015 07:32:32 +0000
Subject: [PATCH] level adjustment for images
---
src/convolutional_layer.c | 52 +++++++++++++++-------------------------------------
1 files changed, 15 insertions(+), 37 deletions(-)
diff --git a/src/convolutional_layer.c b/src/convolutional_layer.c
index ad0d1c1..cd357d3 100644
--- a/src/convolutional_layer.c
+++ b/src/convolutional_layer.c
@@ -29,7 +29,7 @@
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 @@
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)
@@ -61,7 +61,7 @@
layer->biases = calloc(n, sizeof(float));
layer->bias_updates = calloc(n, sizeof(float));
float scale = 1./sqrt(size*size*c);
- for(i = 0; i < c*n*size*size; ++i) layer->filters[i] = scale*rand_normal();
+ for(i = 0; i < c*n*size*size; ++i) layer->filters[i] = 2*scale*rand_uniform() - scale;
for(i = 0; i < n; ++i){
layer->biases[i] = scale;
}
@@ -129,11 +129,10 @@
void backward_bias(float *bias_updates, float *delta, int batch, int n, int size)
{
- float alpha = 1./batch;
int i,b;
for(b = 0; b < batch; ++b){
for(i = 0; i < n; ++i){
- bias_updates[i] += alpha * sum_array(delta+size*(i+b*n), size);
+ bias_updates[i] += sum_array(delta+size*(i+b*n), size);
}
}
}
@@ -167,7 +166,6 @@
void backward_convolutional_layer(convolutional_layer layer, network_state state)
{
- float alpha = 1./layer.batch;
int i;
int m = layer.n;
int n = layer.size*layer.size*layer.c;
@@ -188,7 +186,7 @@
im2col_cpu(im, layer.c, layer.h, layer.w,
layer.size, layer.stride, layer.pad, b);
- gemm(0,1,m,n,k,alpha,a,k,b,k,1,c,n);
+ gemm(0,1,m,n,k,1,a,k,b,k,1,c,n);
if(state.delta){
a = layer.filters;
@@ -202,14 +200,14 @@
}
}
-void update_convolutional_layer(convolutional_layer layer, float learning_rate, float momentum, float decay)
+void update_convolutional_layer(convolutional_layer layer, int batch, float learning_rate, float momentum, float decay)
{
int size = layer.size*layer.size*layer.c*layer.n;
- axpy_cpu(layer.n, learning_rate, layer.bias_updates, 1, layer.biases, 1);
+ axpy_cpu(layer.n, learning_rate/batch, layer.bias_updates, 1, layer.biases, 1);
scal_cpu(layer.n, momentum, layer.bias_updates, 1);
- axpy_cpu(size, -decay, layer.filters, 1, layer.filter_updates, 1);
- axpy_cpu(size, learning_rate, layer.filter_updates, 1, layer.filters, 1);
+ axpy_cpu(size, -decay*batch, layer.filters, 1, layer.filter_updates, 1);
+ axpy_cpu(size, learning_rate/batch, layer.filter_updates, 1, layer.filters, 1);
scal_cpu(size, momentum, layer.filter_updates, 1);
}
@@ -219,42 +217,22 @@
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);
--
Gitblit v1.10.0