SSR: Semi-supervised Soft Rasterizer for single-view 2D to 3D Reconstruction

Accepted at ICCV2021 Workshop [Paper]

0. Download the Dataset

bash scripts/download_dataset.sh

1. Install requirements

pip install -r requirements.txt

and the SoftRas module

git clone https://github.com/ShichenLiu/SoftRas
python setup.py install

2. Train and Validate

python trainval.py -e ubl -sb ../results -d $DATA -r 1

Argument Descriptions:

-e  [Experiment group to run like 'vae' (the rest of the experiment groups are in exp_configs/main_exps.py)] 
-sb [Directory where the experiments are saved]
-r  [Flag for whether to reset the experiments]
-d  [Directory where the datasets are aved]

3. Visualize the Results

Follow these steps to visualize plots. Open results.ipynb, run the first cell to get a dashboard like in the gif below, click on the "plots" tab, then click on "Display plots" or the "images" tab and then click on "Display Images". Parameters of the plots can be adjusted in the dashboard for custom visualizations.

Cite

@article{laradji2021ssr,
  title={SSR: Semi-supervised Soft Rasterizer for single-view 2D to 3D Reconstruction},
  author={Laradji, Issam and Rodr{\'\i}guez, Pau and Vazquez, David and Nowrouzezahrai, Derek},
  journal={arXiv preprint arXiv:2108.09593},
  year={2021}
}