search

yulequan / melanoma-recognition

Repository of paper "Automated Melanoma Recognition in Dermoscopy Images via Very Deep Residual Networks"
http://www.cse.cuhk.edu.hk/~lqyu/skin/
50 stars 23 forks source link

ResNet-50-deploy.prototext has 1000 outputs #2

Closed pjvance closed 7 years ago

pjvance commented 7 years ago

Thanks for sharing!

On the link provided to download a trained network from onedrive, The 'fc1000' layer has 1000 outputs. Is this correct for the trained model?

layer { bottom: "pool5" top: "fc1000" name: "fc1000" type: "InnerProduct" inner_product_param { num_output: 1000 } }

I am attempting to run the model via python and having a little difficulty interpreting the output. Any help would be greatly appreciated.

Many Thanks, Pjvance

yulequan commented 7 years ago

Do you mean the original ResNet-50 network? The original ResNet-50 for ImageNet has 1000 classes and you should fine-tune (only use the previous feature layers) the network.

pjvance commented 7 years ago

Apologies, I had assumed that the pre-trained network provided for download was the network described in the paper. I wished to run the model for myself. So I would need to use the original model and fine-tune using the files provided?

yulequan commented 7 years ago

Yes.