Matching Words for Out -of-distribution Detection

Deep neural networks often exhibit the overconfidence issue when encountering out-of-distribution (00D) samples. To address this, leveraging large-scale pre -trained models like CLIP has shown promise. While CLIP has the capability to encode a vast array of interconnected concepts, current 001) detection methods based on it primarily focus on II) categories and a limited set of OOD categories. In this paper, we propose a novel approach that harnesses the power of WordNet to fully exploit the rich knowledge encapsulated within CLIP, resulting in enhanced OOD detection performance. Our methodology involves constructing a word tree that includes both in-distribution (ID) words and a large set of semantically similar 001) words selected from WordNet. By matching a test image with the concepts of the words in the word tree using CLIP, we estimate the probability of the image being classified as either ID or 00D. Furthermore, we introduce a conditional random field model to effectively handle both the parent-child and the sibling -sibling conflicts in the concept matching results. Extensive experiments under various I1)100D settings demonstrate the effectiveness of our approach and its superiority over state-of-the-art methods.