Delving into Out-of-Distribution Detection with Vision-Language Representations

Authors: Yifei Ming, Ziyang Cai, Jiuxiang Gu, Yiyou Sun, Wei Li, Yixuan Li

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments demonstrate that MCM achieves superior performance on a wide variety of real-world tasks.
Researcher Affiliation Collaboration Yifei Ming1 Ziyang Cai1 Jiuxiang Gu2 Yiyou Sun1 Wei Li3 Yixuan Li1 1Department of Computer Sciences, University of Wisconsin-Madison 2Adobe 3Google Research
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Code is available at https://github.com/ deeplearning-wisc/MCM.
Open Datasets Yes We consider the following ID datasets: CUB-200 [80], STANFORD-CARS [39], FOOD-101 [6], OXFORD-PET [57] and variants of IMAGENET [11].
Dataset Splits No The paper states 'λ is chosen so that a high fraction of ID data (e.g., 95%) is above the threshold' which describes a validation-like process for threshold selection, but it does not specify a formal dataset split (e.g., percentages or sample counts) for this validation.
Hardware Specification No The paper mentions models like CLIP-B/16 and Vi T-B/16, but does not specify the hardware (e.g., GPU/CPU models, memory) used to run the experiments within the provided text.
Software Dependencies No The paper mentions using 'CLIP' and 'Transformer' models but does not list specific software dependencies with version numbers (e.g., Python 3.x, PyTorch 1.x).
Experiment Setup Yes Unless specified otherwise, the temperature is 1 for all experiments.