eli

哇， 你真细心！ `file_list = [] for extension in extensions: file_glob = os.path.join('./tttt', '*.' + extension) file_list.extend(gfile.Glob(file_glob)) file_list Out[11]: ['.\\tttt\\0018_num6235.png', '.\\tttt\\0023_num5141.png', '.\\tttt\\0023_num8005.png', '.\\tttt\\0018_num6235.png', '.\\tttt\\0023_num5141.png', '.\\tttt\\0023_num8005.png']` 这样可能会解决这个重复的问题： ` for file_name in list(set(file_list)):`

i find that in numerical set [1,2,3,4,5,6,7,8,9,0] , this model do not perform very well , is there any further solution to handle this? thanks a lot!

especially between 1 and 7



what slim do in the situation :Filter is larger than the input_featuremap: Filter: (3, 3) Input: (2, 2) can you give a example to explain how it works? thanks a...

i get it ! in tensorflow issues, i find that: Remove unnecessary input-size requirement for convolutions with padding='SAME'