Comprehensive Multi-Modal Interactions for Referring Image Segmentation

Findings (ACL) 2022  ·  Kanishk Jain, Vineet Gandhi ·

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.

PDF Abstract Findings (ACL) 2022 PDF Findings (ACL) 2022 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Referring Expression Segmentation RefCOCOg-val SHNet Overall IoU 49.90 # 12
Referring Expression Segmentation RefCOCO testA SHNet Overall IoU 68.56 # 15
Referring Expression Segmentation RefCOCO+ testA SHNet Overall IoU 58.46 # 14
Referring Expression Segmentation RefCOCO testB SHNet Overall IoU 62.04 # 13
Referring Expression Segmentation RefCOCO+ test B SHNet Overall IoU 44.12 # 14
Referring Expression Segmentation RefCoCo val SHNet Overall IoU 65.32 # 17
Precision@0.9 16.23 # 1
Precision@0.8 46.16 # 1
Precision@0.7 61.21 # 1
Precision@0.6 69.36 # 1
Precision@0.5 75.18 # 1
Referring Expression Segmentation RefCOCO+ val SHNet Overall IoU 52.75 # 16
Referring Expression Segmentation ReferIt SHNet Overall IoU 69.19 # 3

Methods


No methods listed for this paper. Add relevant methods here