Merge pull request #16 from becauseofAI/patch-3

fix the pad bug

darknet2caffe.py

@@ -185,7 +185,7 @@ def cfg2prototxt(cfgfile):
             prev_filters = block['filters']
             convolution_param['kernel_size'] = block['size']
             if block['pad'] == '1':
-                convolution_param['pad'] = str(int(convolution_param['kernel_size'])/2)
+                convolution_param['pad'] = str(int(convolution_param['kernel_size']) // 2)
             convolution_param['stride'] = block['stride']
             if block['batch_normalize'] == '1':
                 convolution_param['bias_term'] = 'false'
@@ -254,7 +254,7 @@ def cfg2prototxt(cfgfile):
             convolution_param['num_output'] = prev_filters
             convolution_param['kernel_size'] = block['size']
             if block['pad'] == '1':
-                convolution_param['pad'] = str(int(convolution_param['kernel_size'])/2)
+                convolution_param['pad'] = str(int(convolution_param['kernel_size']) // 2)
             convolution_param['stride'] = block['stride']
             if block['batch_normalize'] == '1':
                 convolution_param['bias_term'] = 'false'
@@ -320,15 +320,15 @@ def cfg2prototxt(cfgfile):
             pooling_param = OrderedDict()
             pooling_param['stride'] = block['stride']
             pooling_param['pool'] = 'MAX'
-            pooling_param['kernel_size'] = block['size']
-            pooling_param['pad'] = str((int(block['size'])-1)/2)
-            #if (int(block['size']) - int(block['stride'])) % 2 == 0:
-	    #	pooling_param['kernel_size'] = block['size']
-            #	pooling_param['pad'] = str((int(block['size'])-1)/2)
-
-            #if (int(block['size']) - int(block['stride'])) % 2 == 1:
-            #    pooling_param['kernel_size'] = str(int(block['size']) + 1)
-            #	pooling_param['pad'] = str((int(block['size']) + 1)/2)
+            # pooling_param['kernel_size'] = block['size']
+            # pooling_param['pad'] = str((int(block['size'])-1) // 2)
+            if (int(block['size']) - int(block['stride'])) % 2 == 0:
+                pooling_param['kernel_size'] = block['size']
+                pooling_param['pad'] = str((int(block['size'])-1) // 2)
+
+            if (int(block['size']) - int(block['stride'])) % 2 == 1:
+                pooling_param['kernel_size'] = str(int(block['size']) + 1)
+                pooling_param['pad'] = str((int(block['size']) + 1) // 2)
             
             max_layer['pooling_param'] = pooling_param
             layers.append(max_layer)