No abstract

No abstract

Referring image segmentation segments an image from a language expression. With the aim of producing high-quality masks, existing methods often adopt iterative learning approaches that rely on RNNs or stacked attention layers to refine vision-language features. Despite their complexity, RNN-based methods are subject to specific encoder choices, while attention-based methods offer limited gains. In this work, we introduce a simple yet effective alternative for progressively learning discriminative multi-modal features. The core idea of our approach is to leverage a continuously updated query as the representation of the target object and at each iteration, strengthen multi-modal features strongly correlated to the query while weakening less related ones. As the query is initialized by language features and successively updated by object features, our algorithm gradually shifts from being localization-centric to segmentation-centric. This strategy enables the incremental recovery of missing object parts and/or removal of extraneous parts through iteration. Compared to its counterparts, our method is more versatile-it can be plugged into prior arts straightforwardly and consistently bring improvements. Experimental results on the challenging datasets of RefCOCO, RefCOCO+, and G-Ref demonstrate its advantage with respect to the state-of-the-art methods.

Due to the moist environment and inevitable movement, efficient wound closure and healing of vulnerable joint skin remains a great challenge. Herein, a poly(γ‐glutamic acid)‐crosslinked amino‐functionalized PEGylated poly(glycerol sebacate) (γ‐PGA/PEGS‐NH2) adhesive hydrogel is reported. PEGS‐NH2 and γ‐PGA not only forms covalent amide bonds with biological tissue surfaces to achieve strong moist adhesion but also establishes a stable chemically crosslinked network in bulk hydrogels to resist deformation. Furthermore, bioinspired gallic acid‐modified chitosan (CS‐GA) is introduced to enhance moist adhesion via multiple hydrogen bonds and establish a dynamic physically crosslinked network to dissipate energy. Consequently, this adhesive hydrogel strongly adheres to moist biological tissue, showing an adhesion six times higher than that of fibrin glue and comparable to that of strong cyanoacrylate glue. Moreover, benefiting from high mechanical resilience and effective energy dissipation, 200 cycles of loading–unloading mechanical tests conducted under an adhesive state and a full‐thickness rat skin incision model applied on a dynamic nape further confirmed the desirable dynamic tissue adhesion and wound healing performance. Combining the above ideal features with their good injectability and shape‐adaptability to complex contours, such adhesive hydrogels are demonstrated to be promising candidates for joint wound closure and healing in moist and dynamic physiological environment.