Improving Referring Expression Grounding with Cross-modal Attention-guided Erasing
This work improves object localization in images using natural language descriptions, representing an incremental advance in cross-modal alignment methods.
The paper tackles the problem of referring expression grounding by addressing the limitation of previous attention models that focus only on dominant features, proposing a cross-modal attention-guided erasing approach to discover complementary textual-visual correspondences. It achieves state-of-the-art performance on three datasets.
Referring expression grounding aims at locating certain objects or persons in an image with a referring expression, where the key challenge is to comprehend and align various types of information from visual and textual domain, such as visual attributes, location and interactions with surrounding regions. Although the attention mechanism has been successfully applied for cross-modal alignments, previous attention models focus on only the most dominant features of both modalities, and neglect the fact that there could be multiple comprehensive textual-visual correspondences between images and referring expressions. To tackle this issue, we design a novel cross-modal attention-guided erasing approach, where we discard the most dominant information from either textual or visual domains to generate difficult training samples online, and to drive the model to discover complementary textual-visual correspondences. Extensive experiments demonstrate the effectiveness of our proposed method, which achieves state-of-the-art performance on three referring expression grounding datasets.