Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/TRI-ML/sdflabel
Official PyTorch implementation of CVPR 2020 oral "Autolabeling 3D Objects With Differentiable Rendering of SDF Shape Priors"
https://github.com/TRI-ML/sdflabel
Last synced: about 2 months ago
JSON representation
Official PyTorch implementation of CVPR 2020 oral "Autolabeling 3D Objects With Differentiable Rendering of SDF Shape Priors"
- Host: GitHub
- URL: https://github.com/TRI-ML/sdflabel
- Owner: TRI-ML
- License: mit
- Created: 2020-03-29T18:59:36.000Z (over 4 years ago)
- Default Branch: master
- Last Pushed: 2021-10-19T15:04:59.000Z (almost 3 years ago)
- Last Synced: 2024-07-21T21:41:15.287Z (2 months ago)
- Language: Python
- Homepage:
- Size: 10.5 MB
- Stars: 159
- Watchers: 20
- Forks: 20
- Open Issues: 4
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
## Autolabeling 3D Objects With Differentiable Rendering of SDF Shape Priors
Official [PyTorch](https://pytorch.org/) implementation of the CVPR 2020 paper "Autolabeling 3D Objects With Differentiable Rendering of SDF Shape Priors" by the ML Team at [Toyota Research Institute (TRI)](https://www.tri.global/), cf. [References](#references) below.
[**[Full paper]**](https://arxiv.org/pdf/1911.11288.pdf) [**[YouTube]**](https://www.youtube.com/watch?v=Utzj-kfWHP4)## Setting up your environment
To set up the environment using conda, use the following commands:
```
conda env create -n sdflabel -f environment.yml
conda activate sdflabel
```
Add the sdfrenderer directory to *PYTHONPATH*:
```
export PYTHONPATH="${PYTHONPATH}:/path/to/sdfrenderer"
```## Optimization demo
To run the optimization demo, first download the [data folder](https://drive.google.com/file/d/1cvLeXDhaghjzCK-gmnQbxh8rTvd71miw/view?usp=sharing).
Then, extract the archive to the root folder of the project and run the following command:
```
python main.py configs/config_refine.ini --demo
```## Training CSS network
To train the CSS network, run the following command:
```
python main.py configs/config_train.ini --train
```### Dataset format
The dataset of crops represents a collection of detected *RGB patches (CSS input)*, corresponding *NOCS patches (CSS output)*, and a JSON DB file comprising the patch relevant information (most importantly *SDF latent vectors* corresponding to the depicted 3D models).
An example of such dataset is located in the *data/db/crops* folder.## Optimization
Download [KITTI 3D](http://www.cvlibs.net/datasets/kitti/eval_object.php?obj_benchmark=3d) and modify the *kitti_path* in the config file *config_refine.ini* accordingly.
To run optimization on the KITTI 3D dataset, run the following command:
```
python main.py configs/config_refine.ini --refine
```
Upon completion, autolabels will be stored to the *output* folder specified in the config file (*output* -> *labels*).
To evaluate the generated dump, run:
```
python main.py configs/config_refine.ini --evaluate
```## License
The source code is released under the [MIT license](LICENSE.md).
## References
#### Autolabeling 3D Objects With Differentiable Rendering of SDF Shape Priors (CVPR 2020 oral)
*Sergey Zakharov\*, Wadim Kehl\*, Arjun Bhargava, Adrien Gaidon*```
@inproceedings{sdflabel,
author = {Sergey Zakharov and Wadim Kehl and Arjun Bhargava and Adrien Gaidon},
title = {Autolabeling 3D Objects with Differentiable Rendering of SDF Shape Priors},
booktitle = {IEEE Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}
```