Home

Awesome

SPSG

SPSG presents a self-supervised approach to generate high-quality, colored 3D models of scenes from RGB-D scan observations by learning to infer unobserved scene geometry and color. Rather than relying on 3D reconstruction losses to inform our 3D geometry and color reconstruction, we propose adversarial and perceptual losses operating on 2D renderings in order to achieve high-resolution, high-quality colored reconstructions of scenes. For more details, please see our paper SPSG: Self-Supervised Photometric Scene Generation from RGB-D Scans.

<img src="spsg.jpg">

Code

Installation:

Training is implemented with PyTorch. This code was developed under PyTorch 1.2.0 and Python 2.7.

Please compile the extension modules by running the install_utils.sh script.

Training:

Testing

Data:

Citation:

If you find our work useful in your research, please consider citing:

@inproceedings{dai2021spsg,
    title={SPSG: Self-Supervised Photometric Scene Generation from RGB-D Scans},
    author={Dai, Angela and Siddiqui, Yawar and Thies, Justus and Valentin, Julien and Nie{\ss}ner, Matthias},
	booktitle={Proc. Computer Vision and Pattern Recognition (CVPR), IEEE},
	year={2021}
}