In vision-language models (VLMs), visual tokens usually consume a significant amount of computational overhead, despite their sparser information density compared to text tokens. To address this, ...
Abstract: Large Vision-Language Models (LVLMs) suffer from severe object hallucinations, leading them to frequently generate outputs that do not correspond to the image content, significantly reducing ...
Abstract: Edge intelligent visual monitoring of substations performs intelligent detection on collected images at edge intelligent perception terminals, representing a novel substation intelligent ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results