Robust Saliency-Aware Distillation for Few-shot Fine-grained Visual Recognition

arXiv:2305.07180v3 Announce Type: replace
Abstract: Recognizing novel sub-categories with scarce samples is an essential and challenging research topic in computer vision. Existing literature addresses this challenge by employing local-based representation approaches, which may not sufficiently facilitate meaningful object-specific semantic understanding, leading to a reliance on apparent background correlations. Moreover, they primarily rely on high-dimensional local descriptors to construct complex embedding space, potentially limiting the generalization. To address the above challenges, this article proposes a novel model, Robust Saliency-aware Distillation (RSaD), …

abstract arxiv challenge computer computer vision correlations cs.cv distillation few-shot fine-grained literature novel recognition reliance representation research robust samples semantic type understanding vision visual