Dynamic Relation Inference via Verb Embeddings
This addresses a specific bottleneck in vision-language models for tasks requiring relational understanding, with incremental but practical gains.
The paper tackles CLIP's limitation in inferring relationships among objects in images by proposing DRIVE, which fine-tunes CLIP with hard negatives and a novel loss function, resulting in significant improvements in zero-shot relation inference accuracy over CLIP and state-of-the-art models.
CLIP has demonstrated exceptional image-text matching capabilities due to its training on contrastive learning tasks. Past research has suggested that whereas CLIP effectively matches text to images when the matching can be achieved just by matching the text with the objects in the image, CLIP struggles when the matching depends on representing the relationship among the objects in the images (i.e., inferring relations). Previous attempts to address this limitation by training CLIP on relation detection datasets with only linguistic supervision have met with limited success. In this paper, we offer insights and practical methods to advance the field of relation inference from images. This paper approaches the task of creating a model that effectively detects relations among the objects in images by producing text and image embeddings that capture relationships through linguistic supervision. To this end, we propose Dynamic Relation Inference via Verb Embeddings (DRIVE), which augments the COCO dataset, fine-tunes CLIP with hard negatives subject-relation-object triples and corresponding images, and introduces a novel loss function to improve relation detection. Evaluated on multiple CLIP-based models, our method significantly improves zero-shot relation inference accuracy in both frozen and fine-tuned settings, significantly outperforming CLIP and state-of-the-art models while generalizing well on unseen data.