convolution_neural_network.py 文件源码

def pooling(self,featuremaps,size_pooling,type='average_pool'):
        #pooling process
        size_map = len(featuremaps[0])
        size_pooled = int(size_map/size_pooling)
        featuremap_pooled = []
        for i_map in range(len(featuremaps)):
            map = featuremaps[i_map]
            map_pooled = []
            for i_focus in range(0,size_map,size_pooling):
                for j_focus in range(0, size_map, size_pooling):
                    focus = map[i_focus:i_focus + size_pooling, j_focus:j_focus + size_pooling]
                    if type == 'average_pool':
                        #average pooling
                        map_pooled.append(np.average(focus))
                    elif type == 'max_pooling':
                        #max pooling
                        map_pooled.append(np.max(focus))
            map_pooled = np.asmatrix(map_pooled).reshape(size_pooled,size_pooled)
            featuremap_pooled.append(map_pooled)
        return featuremap_pooled