3D Shape Retrieval

3D shape retrieval aims to efficiently find similar 3D models within large databases, using various input modalities like single or multi-view images, 3D sketches (even from VR environments), or even partial scans. Current research emphasizes robust methods handling occlusions, unseen objects, and noisy data, employing techniques like keypoint-based embeddings, adaptive attention mechanisms across multiple views, and part-based aggregation networks to improve retrieval accuracy. This field is crucial for applications ranging from CAD model searching to augmented and virtual reality experiences, driving advancements in both computer vision and 3D modeling.