WALT3D: Generating Realistic Training Data from Time-Lapse Imagery for Reconstructing Dynamic Objects under Occlusion

arXiv:2403.19022v1 Announce Type: new
Abstract: Current methods for 2D and 3D object understanding struggle with severe occlusions in busy urban environments, partly due to the lack of large-scale labeled ground-truth annotations for learning occlusion. In this work, we introduce a novel framework for automatically generating a large, realistic dataset of dynamic objects under occlusions using freely available time-lapse imagery. By leveraging off-the-shelf 2D (bounding box, segmentation, keypoint) and 3D (pose, shape) predictions as pseudo-groundtruth, unoccluded 3D objects are identified automatically …

3d object abstract annotations arxiv cs.cv current data dynamic environments framework ground-truth novel object objects scale struggle training training data truth type understanding urban work