From 0f645836f193e75c4c3b718369e6fab15b5d19c5 Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Wed, 11 Feb 2015 03:41:03 +0000
Subject: [PATCH] Detection is back, baby\!
---
src/crop_layer.c | 56 ++++++++++----------------------------------------------
1 files changed, 10 insertions(+), 46 deletions(-)
diff --git a/src/crop_layer.c b/src/crop_layer.c
index 2a5007a..3f0011d 100644
--- a/src/crop_layer.c
+++ b/src/crop_layer.c
@@ -1,4 +1,5 @@
#include "crop_layer.h"
+#include "cuda.h"
#include <stdio.h>
image get_crop_image(crop_layer layer)
@@ -22,19 +23,24 @@
layer->crop_height = crop_height;
layer->output = calloc(crop_width*crop_height * c*batch, sizeof(float));
#ifdef GPU
- layer->output_cl = cl_make_array(layer->output, crop_width*crop_height*c*batch);
+ layer->output_gpu = cuda_make_array(layer->output, crop_width*crop_height*c*batch);
#endif
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 dw = rand()%(layer.w - layer.crop_width);
+ int dh = rand()%(layer.h - layer.crop_height + 1);
+ 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){
@@ -53,45 +59,3 @@
}
}
-#ifdef GPU
-cl_kernel get_crop_kernel()
-{
- static int init = 0;
- static cl_kernel kernel;
- if(!init){
- kernel = get_kernel("src/crop_layer.cl", "forward", 0);
- init = 1;
- }
- return kernel;
-}
-
-void forward_crop_layer_gpu(crop_layer layer, cl_mem input)
-{
- int flip = (layer.flip && rand()%2);
- int dh = rand()%(layer.h - layer.crop_height);
- int dw = rand()%(layer.w - layer.crop_width);
- int size = layer.batch*layer.c*layer.crop_width*layer.crop_height;
-
- cl_kernel kernel = get_crop_kernel();
- cl_command_queue queue = cl.queue;
-
- cl_uint i = 0;
- cl.error = clSetKernelArg(kernel, i++, sizeof(input), (void*) &input);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.c), (void*) &layer.c);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.h), (void*) &layer.h);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.w), (void*) &layer.w);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.crop_height), (void*) &layer.crop_height);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.crop_width), (void*) &layer.crop_width);
- cl.error = clSetKernelArg(kernel, i++, sizeof(dh), (void*) &dh);
- cl.error = clSetKernelArg(kernel, i++, sizeof(dw), (void*) &dw);
- cl.error = clSetKernelArg(kernel, i++, sizeof(flip), (void*) &flip);
- cl.error = clSetKernelArg(kernel, i++, sizeof(layer.output_cl), (void*) &layer.output_cl);
- check_error(cl);
-
- const size_t global_size[] = {size};
-
- cl.error = clEnqueueNDRangeKernel(queue, kernel, 1, 0, global_size, 0, 0, 0, 0);
- check_error(cl);
-}
-
-#endif
--
Gitblit v1.10.0