
This paper addresses the challenges in object 6D pose estimation caused by illumination changes, proposing an improved Gen6D method for robust operation under high dynamic ambient light conditions.
Jan 1, 2026

SFRF6D is a model-free 6D pose estimation framework that needs no CAD model and only a few RGB-D reference images with known poses, combining reference-view rendering, geometry-guided sparse cross-view attention and coarse-to-fine pose selection for robustness under occlusion.
Jan 1, 2025

GCMA6D is an RGB-D 6D pose estimation network that adds 3D graph convolution to the point-cloud branch, Large Kernel Attention to the image branch, and cross-modality attention with Squeeze-and-Excitation reweighting to fuse the two, for occluded and low-texture objects.
Jan 1, 2025