Deep Object Occlusion Relationship Detection Framework Based on Associative Embedding Clustering
Peiyong Gong, Kai Zheng, Ting Liu, Huixuan Zhao, Yi Jiang · 2025
Visual spatial relationship detection is essential for understanding scenes in images, focusing on detecting objects and recognizing spatial relationships between object pairs. However, occlusion, a critical visual spatial relationship and semantic feature, has been insufficiently explored. To address this issue, we propose a pioneering approach termed DOORD-AEC, specifically designed for detecting occlusion spatial relationships among objects. DOORD-AEC algorithm introduces associative embedding clustering to supervise a convolutional neural network with two branches, enabling it to take in an input image and produce a triplet set representing occlusion spatial relationships. The network learns to simultaneously identify all the targets and occlusions that make up the triplet set and group them together using associative embedding clustering. Experiments conducted on the KITTI-based dataset demonstrate the effectiveness of our method.