mxnet-model-gallery
mxnet-model-gallery copied to clipboard
Convert Inception21k to PyTorch
Hi,
I am trying to convert the Inception21k model to PyTorch. I used the MMdnn tool to do the conversion; however, I am only seeing 65% top1 accuracy on the validation set (below the reported 68%).
What was the image preprocessing used to train the network? Currently I am doing:
- Resize shortest size of image to 256
- Center crop to 224
- Multiply by 255 to scale image input range to [0, 255]
- Subtract mean of 117 from the image
Is this correct?