/Confident_classifier

Training Confidence-Calibrated Classifier for Detecting Out-of-Distribution Samples / ICLR 2018

Primary LanguagePythonMIT LicenseMIT

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:

Training scripts

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