SMohammadi89 / PointView-GCN

The code and dataset will be available soon here
MIT License
59 stars 17 forks source link

Code for PointView-GCN [ICIP2021].

Seyed Saber Mohammadi, Yiming Wang, Alessio Del Bue. PointView-GCN: 3D shape classification with multi-view point clouds. You can find IEEE version of the paper here.

Citation

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

@inproceedings{mohammadi2021pointview,
  title={Pointview-GCN: 3D Shape Classification With Multi-View Point Clouds},
  author={Mohammadi, Seyed Saber and Wang, Yiming and Del Bue, Alessio},
  booktitle={2021 IEEE International Conference on Image Processing (ICIP)},
  pages={3103--3107},
  year={2021},
  organization={IEEE}
}

Dataset

You can find our dataset with partial single-view PCDs generated from benchmark dataset ModelNet40. Plese download the dataset, creat a directory named "single_view_modelnet" and put it under "data" directory.

Training

First use the pre-trained model to extract the features from each single-view PCD:

cd Feature_extraction
python main.py

Then apply the GCN to aggregate and classify the features:

cd GCN
python main.py

You can also train the backbone from scratch:

cd PointNet++
python main.py

Dataset generation

First download the normalize version of ModelNet40 dataset ModelNet40_normalized and put it under the "data" directory. then run the following comment:

cd dataset_rendering
python dataset_capturing.py --out-split-dir /train/ && python dataset_capturing.py --out-split-dir /test/

Note that, since the dataset generation takes a huge amount of the time, we provided the final version of the generated single-view PCDs.