DIVE: Towards Descriptive and Diverse Visual Commonsense Generation
This addresses a gap in visual commonsense generation for AI systems aiming for human-level visual understanding, though it appears incremental as it builds on existing resources and objectives.
The paper tackles the problem of generating descriptive and diverse visual commonsense inferences, which current research overlooks, by proposing the DIVE framework with generic inference filtering and contrastive retrieval learning, achieving human-level performance on Visual Commonsense Graphs and outperforming state-of-the-art models.
Towards human-level visual understanding, visual commonsense generation has been introduced to generate commonsense inferences beyond images. However, current research on visual commonsense generation has overlooked an important human cognitive ability: generating descriptive and diverse inferences. In this work, we propose a novel visual commonsense generation framework, called DIVE, which aims to improve the descriptiveness and diversity of generated inferences. DIVE involves two methods, generic inference filtering and contrastive retrieval learning, which address the limitations of existing visual commonsense resources and training objectives. Experimental results verify that DIVE outperforms state-of-the-art models for visual commonsense generation in terms of both descriptiveness and diversity, while showing a superior quality in generating unique and novel inferences. Notably, DIVE achieves human-level descriptiveness and diversity on Visual Commonsense Graphs. Furthermore, human evaluations confirm that DIVE aligns closely with human judgments on descriptiveness and diversity\footnote{Our code and dataset are available at https://github.com/Park-ing-lot/DIVE.