This is a repository for the paper Cross-Dimensional Refined Learning for Real-Time 3D Visual Perception from Monocular Video.

python valscene_inference.py

The key modes can be configured under configs/inference.yaml, where disabling MODEL.DEPTH_PREDICTION release the model into the geometric-semantic inference mode. The geometric-semantic information has been learned by MAP optimization with the help of 2D priors.

Please consider citing our paper and give a ⭐ if you find this repository useful.

@inproceedings{hong2023cross,
    author    = {Hong, Ziyang and Yue, C. Patrick},
    title     = {Cross-Dimensional Refined Learning for Real-Time 3D Visual Perception from Monocular Video},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
    month     = {October},
    year      = {2023},
    pages     = {2169-2178}
}