/nn-tiny-imagenet-200

43% accuracy on tiny-imagenet-200

Primary LanguagePythonMIT LicenseMIT

NN for tiny-imagenet-200

  • tiny-imagenet-200 (64x64xRGB)
  • No pretrained weights
  • 43% accuracy (ratio)

Env notes:

  • conda create -n NAME python=3.6
  • conda install scikit-learn
  • conda install pytorch=0.4.1 cuda90 -c pytorch
  • conda install torchvision

Expected typical output:

Device: cuda

CUDA Version: 9.0.176

Downloading tiny-imagenet-200... ./tiny-imagenet-200.zip

Done.

Weight shapes: [torch.Size([64, 3, 3, 3]), torch.Size([64]), torch.Size([64]), torch.Size([128, 64, 3, 3]), torch.Size([128]), torch.Size([128]), torch.Size([128, 128, 3, 3]), torch.Size([128]), torch.Size([128]), torch.Size([256, 128, 3, 3]), torch.Size([256]), torch.Size([256]), torch.Size([256, 256, 3, 3]), torch.Size([256]), torch.Size([256]), torch.Size([512, 256, 3, 3]), torch.Size([512]), torch.Size([512]), torch.Size([200, 8192]), torch.Size([200])]

Calculating weights... Done.

================================================== Epoch #0 | 14.97% accuracy ================================================== Epoch #1 | 20.82% accuracy ================================================== Epoch #2 | 25.86% accuracy ================================================== Epoch #3 | 30.29% accuracy ================================================== Epoch #4 | 31.89% accuracy ================================================== Epoch #5 | 33.99% accuracy ================================================== Epoch #6 | 35.83% accuracy ================================================== Epoch #7 | 37.82% accuracy ================================================== Epoch #8 | 37.72% accuracy ================================================== Epoch #9 | 40.07% accuracy ================================================== Epoch #10 | 40.96% accuracy ================================================== Epoch #11 | 42.01% accuracy ================================================== Epoch #12 | 42.41% accuracy ================================================== Epoch #13 | 42.55% accuracy ================================================== Epoch #14 | 43.19% accuracy ================================================== Epoch #15 | 43.00% accuracy ================================================== Epoch #16 | 44.09% accuracy ================================================== Epoch #17 | 44.37% accuracy ================================================== Epoch #18 | 45.24% accuracy ================================================== Epoch #19 | 44.41% accuracy ================================================== Epoch #20 | 44.41% accuracy

Time is against us!

Train: 71.58% accuracy

Val: 44.13% accuracy

Test: 44.74% accuracy