CL-CLIP: CLIP-Based Continual Learning Framework with Cost-Volume Category Decoupling for Object Detection
Researchers have developed CL-CLIP, a new framework for continual object detection that leverages CLIP's vision-language capabilities. This approach aims to enable object detectors to learn new categories over time without forgetting previously acquired knowledge. CL-CLIP utilizes a cost-volume guided category decoupling method to process visual tokens and class text embeddings, improving performance on datasets like PASCAL VOC and MS-COCO. AI
IMPACT Enhances the ability of AI models to learn new visual categories over time without catastrophic forgetting.