Comprehensive Multi-Modal Interactions for Referring Image Segmentation
This work addresses the challenge of accurately segmenting images based on natural language descriptions, which is important for applications in computer vision and human-computer interaction, representing an incremental improvement over existing methods.
The paper tackles the problem of referring image segmentation by proposing a method that simultaneously handles cross-modal and intra-modal interactions, achieving considerable performance gains over state-of-the-art methods on four benchmark datasets.
We investigate Referring Image Segmentation (RIS), which outputs a segmentation map corresponding to the natural language description. Addressing RIS efficiently requires considering the interactions happening across visual and linguistic modalities and the interactions within each modality. Existing methods are limited because they either compute different forms of interactions sequentially (leading to error propagation) or ignore intramodal interactions. We address this limitation by performing all three interactions simultaneously through a Synchronous Multi-Modal Fusion Module (SFM). Moreover, to produce refined segmentation masks, we propose a novel Hierarchical Cross-Modal Aggregation Module (HCAM), where linguistic features facilitate the exchange of contextual information across the visual hierarchy. We present thorough ablation studies and validate our approach's performance on four benchmark datasets, showing considerable performance gains over the existing state-of-the-art (SOTA) methods.