VT-CLIP: Enhancing Vision-Language Models with Visual-guided Texts [article]

Longtian Qiu, Renrui Zhang, Ziyu Guo, Ziyao Zeng, Zilu Guo, Yafeng Li, Guangnan Zhang
2023 arXiv   pre-print
Contrastive Language-Image Pre-training (CLIP) has drawn increasing attention recently for its transferable visual representation learning. However, due to the semantic gap within datasets, CLIP's pre-trained image-text alignment becomes sub-optimal on downstream tasks, which severely harms its transferring performance. To better adapt the cross-modality embedding space, we propose to enhance CLIP via Visual-guided Texts, named VT-CLIP. Specifically, we guide textual features of different
more » ... ries to adaptively explore informative regions on the image and aggregate visual features by attention mechanisms. In this way, the texts become visual-guided, namely, more semantically correlated with downstream images, which greatly benefits the category-wise matching process. In few-shot settings, we evaluate our VT-CLIP on 11 well-known classification datasets to demonstrate its effectiveness.
arXiv:2112.02399v3 fatcat:rtd7owp2tjablka7rldnjidz7q