Training Confidence-Calibrated Classifier for Detecting Out-of-Distribution Samples

This project is for the paper "Training Confidence-Calibrated Classifier for Detecting Out-of-Distribution Samples". Some codes are from odin-pytorch.

Preliminaries

It is tested under Ubuntu Linux 16.04.1 and Python 2.7 environment, and requries Pytorch package to be installed:

• Pytorch: Only GPU version is available.

Downloading Out-of-Distribtion Datasets

We use download links of two out-of-distributin datasets from odin-pytorch:

• Tiny-ImageNet (resize)
• LSUN (resize)

Training scripts

• run_cross_entropy.sh: train the models using standard cross entropy loss.
• run_joint_confidence.sh: train the models using joint confidence loss.

Test scripts

• test.sh --dataset --out_dataset --pre_trained_net
  --dataset = name of in-distribution (svhn or cifar10)
  --out_dataset = name of out-of-distribution (svhn, cifar10, lsun or imagenet)
  --pre_trained_net = path to pre_trained_net