search

yangerkun / IJCAI2018_SSDH

Semantic Structure-based Unsupervised Deep Hashing IJCAI2018
42 stars 15 forks source link

CIFAR10 input image dimensions? #4

Open codtiger opened 4 years ago

codtiger commented 4 years ago

So some changes was done to the code 4 days ago by the author, where he seems to have replaced cifar10 text files pointing to database, query and train images with MFlickr Dataset. Should the cifar 10 images still be resized to 224*224 and then fed to the VGG network or it is safe to change the placeholder to 64x64?

yangerkun commented 4 years ago

So some changes was done to the code 4 days ago by the author, where he seems to have replaced cifar10 text files pointing to database, query and train images with MFlickr Dataset. Should the cifar 10 images still be resized to 224*224 and then fed to the VGG network or it is safe to change the placeholder to 64x64?

I update the code to make it more clear, our paper doesn't test performance on Cifar10, so I provide the .txt file for flickr dataset. You can change to other datasets accordingly. Since VGG network only accept 224x224 inputs, so your still need to resize them to 224x224.

codtiger commented 4 years ago

Thank you for the response.