Affinity Attention Graph Neural Network for Weakly Supervised Semantic Segmentation.

Area of Science:

• Computer Vision
• Machine Learning
• Artificial Intelligence

Background:

• Weakly supervised semantic segmentation offers reduced human annotation costs.
• Bounding box supervision presents a challenge for training accurate semantic segmentation models.

Purpose of the Study:

• To develop an effective method for bounding box supervised semantic segmentation.
• To propose a novel graph neural network architecture for semantic label propagation.

Main Methods:

• Proposed the Affinity Attention Graph Neural Network (A²GNN) model.
• Generated semantic-aware seeds and formed semantic graphs using a novel affinity Convolutional Neural Network (CNN).
• Introduced a new loss function and consistency-checking mechanism to leverage bounding box constraints.

Main Results:

• Achieved state-of-the-art performance on the Pascal VOC 2012 dataset (76.5% val, 75.2% test).
• Demonstrated applicability to bounding box supervised instance segmentation and other weakly supervised tasks.
• Obtained state-of-the-art or comparable performance on PASCAL VOC and COCO datasets.

Conclusions:

• The A²GNN effectively addresses the challenges of bounding box supervised semantic segmentation.
• The proposed method offers a cost-effective and high-performance solution for semantic segmentation tasks.
• The approach shows broad applicability across various weakly supervised segmentation scenarios.