Classifier-head Informed Feature Masking and Prototype-based Logit Smoothing for Out-of-Distribution Detection

no code implementations • 27 Oct 2023 • Zhuohao Sun, Yiqiao Qiu, Zhijun Tan, Weishi Zheng, Ruixuan Wang

Logit smoothing computes the cosine similarity between the feature vector of the test sample and the prototype of the predicted ID class at the penultimate layer and uses the similarity as an adaptive temperature factor on the logit to alleviate the network's overconfidence prediction for OOD data.

Out-of-Distribution Detection Out of Distribution (OOD) Detection