Detecting Out-of-distribution Data through In-distribution Class Prior

Authors: Xue Jiang, Feng Liu, Zhen Fang, Hong Chen, Tongliang Liu, Feng Zheng, Bo Han

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that both strategies can improve the OOD detection performance when the ID model is pre-trained with imbalanced data, reflecting the importance of ID-class prior in OOD detection.
Researcher Affiliation Academia 1Department of Computer Science and Engineering, Southern University of Science and Technology 2Department of Computer Science, Hong Kong Baptist University 3University of Melbourne 4Australian Artificial Intelligence Institute, University of Technology Sydney 5College of Informatics, Huazhong Agricultural University 6Mohamed bin Zayed University of Artificial Intelligence 7Sydney AI Centre, The University of Sydney.
Pseudocode No The paper does not contain clearly labeled algorithm sections or code-like formatted procedures (pseudocode).
Open Source Code Yes The codes are available at https://github. com/tmlr-group/class_prior.
Open Datasets Yes We construct a series of imbalanced ID datasets that are sampled by the Pareto distribution in Image Net-1K dataset. (Deng et al., 2009) ... the OOD datasets include the subsets of i Naturalist (Horn et al., 2018), SUN (Xiao et al., 2010), Places (Zhou et al., 2018), and Textures (Cimpoi et al., 2014).
Dataset Splits No The paper discusses training and test data, but it does not provide specific train/validation/test dataset splits (e.g., percentages, sample counts, or explicit references to predefined splits for reproduction).
Hardware Specification Yes Note that, all methods are realized by Pytorch 1.6.0 with CUDA 10.2, where we use several NVIDIA Tesla V100 GPUs.
Software Dependencies Yes We use mmclassification3 (Contributors, 2020) with Apache-2.0 license to train ID models. Note that, all methods are realized by Pytorch 1.6.0 with CUDA 10.2, where we use several NVIDIA Tesla V100 GPUs.
Experiment Setup No The training details of Res Net (He et al., 2016) and Mobile Net (Howard et al., 2019) follow the default setting in mmclassification.