Closed chinakook closed 6 years ago
fashion-mnist benchmark
INFO:__main__:Training on 20 images...
INFO:__main__:Testing on 20 images...
Total test accuracy = 0.6
INFO:__main__:Training on 100 images...
INFO:__main__:Testing on 20 images...
Total test accuracy = 0.7
INFO:__main__:Training on 1000 images...
INFO:__main__:Testing on 10000 images...
Total test accuracy = 0.7235
@StephenZengCn But my score on fashion-mnist is:
INFO:__main__:Training on 20 images...
INFO:__main__:Testing on 20 images...
Total test accuracy = 0.15
INFO:__main__:Training on 100 images...
INFO:__main__:Testing on 20 images...
Total test accuracy = 0.05
@StephenZengCn I've got wrong data ealier. Now,I have result as follows:
INFO:__main__:Training on 20 images...
INFO:__main__:Testing on 20 images...
Total test accuracy = 0.55
INFO:__main__:Training on 100 images...
INFO:__main__:Testing on 20 images...
Total test accuracy = 0.5
The result is not ideal.
It's more general and complicated than mnist. https://github.com/zalandoresearch/fashion-mnist