Open Franklin-Yao opened 4 years ago
Hi nice job. I try to reproduce the MinImageNet dataset , too. I get some results better than the author's. but I did not change the net's structure. that make me confused. more import , the result is not stable in 20 way n shot. could you communicate with me about the how the situation about the verification on the MinImageNet dataset ? is it different with the Omniglot ?I want to fingre out what's wrong I had made in my work.
Hi guys,
I got a updated implementation which acheived over 79% acc on the MiniImageNet 5-shot task. It may be helpful for some guys in 2020. https://github.com/Franklin-Yao/StrongProtoNet