search

kkanshul / finegan

FineGAN: Unsupervised Hierarchical Disentanglement for Fine-grained Object Generation and Discovery
http://krsingh.cs.ucdavis.edu/krishna_files/papers/finegan/index.html
BSD 2-Clause "Simplified" License
277 stars 43 forks source link

About fine-tuned inception model file #7

Closed bravotty closed 5 years ago

bravotty commented 5 years ago

Hi! When i test your fine-tuned model file, it happens some errors: Error(s) in loading state_dict for Inception3: Missing key(s) in state_dict: "AuxLogits.fc.weight", "AuxLogits.fc.bias", "fc.weight", "fc.bias". Unexpected key(s) in state_dict: "fc_new.weight", "fc_new.bias", "AuxLogits.fc_new.weight", "AuxLogits.fc_new.bias". Could u mind send a fine-tuned inception model file(inceptionv2.py) for me, thx!!!

Other Question~, how to fine-tuned the inception model on bird datasets?

utkarshojha commented 5 years ago

Hi! Thanks for pointing that out. I've added the new inception model here (https://github.com/kkanshul/finegan/blob/master/code/inception.py). Right now it is for birds, and if you need it for some other dataset, just change the last fully connected layer dimension (200 for cars, 120 for dogs).

As for the fine tuning of inception net, I just initialize the layers (except the last fully connected layer) with the inception network trained on ImageNet dataset. I then further train the network on the complete CUB dataset. I create a 80-20 split of CUB and use the 20 split as validation set to determine when to stop training the model.

Thanks ~

bravotty commented 5 years ago

thx!