evaluation part have a tensor size bug..

[Noumanijaz744]

@yangdongchao its a great work .i have used the baseline extracted features and trained the model the accuracy during training is good and got the best accuracy of 97.8% .while doing the evaluation i am getting the issue of the calculated output size which is so small .as can be seen in the picture.

[Noumanijaz744]

please help me in solving this issue i used the same features used in the baseline implementation.