HNC: Leveraging Hard Negative Captions towards Models with Fine-Grained Visual-Linguistic Comprehension Capabilities
For the vision-language community, this work addresses the weak association in web-collected image-text pairs by providing a method to enhance fine-grained comprehension, though it is incremental as it builds on existing ITM frameworks.
The paper introduces Hard Negative Captions (HNC), an automatically generated dataset of hard negative captions for Image-Text Matching (ITM) training, to improve fine-grained cross-modal understanding. Models trained on HNC show improved zero-shot mismatch detection and robustness to noisy visual inputs, with comparable or better fine-tuning initialization.
Image-Text-Matching (ITM) is one of the defacto methods of learning generalized representations from a large corpus in Vision and Language (VL). However, due to the weak association between the web-collected image-text pairs, models fail to show a fine-grained understanding of the combined semantics of these modalities. To address this issue we propose Hard Negative Captions (HNC): an automatically created dataset containing foiled hard negative captions for ITM training towards achieving fine-grained cross-modal comprehension in VL. Additionally, we provide a challenging manually-created test set for benchmarking models on a fine-grained cross-modal mismatch task with varying levels of compositional complexity. Our results show the effectiveness of training on HNC by improving the models' zero-shot capabilities in detecting mismatches on diagnostic tasks and performing robustly under noisy visual input scenarios. Also, we demonstrate that HNC models yield a comparable or better initialization for fine-tuning