X-Decoder is designed to be highly versatile, supporting tasks at various granularities without needing task-specific architectures: Universal Image Segmentation : Seamlessly handles Semantic, Instance, and Panoptic Segmentation Vision-Language Integration : Supports Image Captioning Image-Text Retrieval Visual Question Answering (VQA) Referring Tasks : Excels at Referring Segmentation
Copyright 2012 Joel Nielsen |