Abstract: Visual grounding tasks aim to localize image regions based on natural language references. In this work, we ex-plore whether generative VLMs predominantly trained on image-text data could be ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback