CIFAR-10 bad results, loss: 1.1991, Accuracy: 6296/11001 (57%)

1

Hi,
Can someone please take a look and help me understand what am I doing wrong?
Original training set (https://www.kaggle.com/c/cifar-10/data) was split into train and validation sets.
Everything is documented here:

https://github.com/QuantScientist/Deep-Learning-Boot-Camp/blob/master/day%2002%20PyTORCH%20and%20PyCUDA/PyTorch/21-PyTorch-CIFAR-10-Custom-data-loader-from-scratch.ipynb

Thanks,

2

The network is too simple.
Try some complex and successful networks such as vgg and resnet.
Or make your network deeper.
And larger epochs(e,g epochs:100) must should be taken.
Actually, there are lots of things you can do to improve accuracy.

1 Like
3

Many Thanks,
just wanted to know its not my own Dataset loader that is causing this issue, rather it is the simple network.