Enhanced Back-Projection of Vision Features for 3D Symmetry Detection

WACV 2026 — Work in Progress

This repository is actively being prepared for release. Additional code modules and the associated datasets will be uploaded soon to accompany the accepted paper.

Project page: https://spulp.github.io/EnhancedBackProjection/

Important

Environment Setup

• Python Version: 3.12
• It is recommended to use a virtual environment for package management.
• Requires GPU with CUDA 12.9.

Creating a Virtual Environment

Using venv

python -m venv p12
source p12/bin/activate

Using Miniconda

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm ~/miniconda3/miniconda.sh
source ~/miniconda3/bin/activate
conda init --all

conda create -n p12 python=3.12
conda install -n p12 pip
conda activate p12

Installing Required Libraries

All required libraries are listed in requirements.txt. Install them with:

pip install -r requirements.txt

Or install them individually:

pip3 install torch torchvision --index-url https://download.pytorch.org/whl/cu129
pip install --extra-index-url https://miropsota.github.io/torch_packages_builder pytorch3d==0.7.8+pt2.8.0cu129
pip install point_cloud_utils
pip install torch_geometric
pip install torch-cluster -f https://data.pyg.org/whl/torch-2.8.0+cu129.html
pip install trimesh pyvista scikit-learn

References

This code is inspired by the following repositories:

• Back-to-3D Few-Shot Keypoints
• COPS

Example Usage

To run the main script, use:

main.py <path_to_obj_file> <planes_or_axes> <mode>

• <path_to_obj_file>: Path to an .obj file (e.g., objects/example_planes.obj)
• <planes_or_axes>: Specify the planes or axes to use (e.g., planes)
• <mode>: Choose one of the following modes:
  • FM: Feature-Mesh sampling
  • RM: Raw-Mesh
  • PC: Point Cloud

Example:

main.py objects/example_planes.obj planes FM
main.py objects/example_axes.obj axes RM

Notes on Code Structure and Usage

The code does not provide extensive explanations for each part, but with enough digging (e.g., looking at function names and structure), you can deduce the workflow. The entry point is the main.py script, which is straightforward to follow.

When executing, a visual interface pops up. The console will display commands for interacting with the viewer, summarized here:

Note: There is a known bug when using the PC mode—the point clouds in the viewer appear larger than normal. However, the computed planes can still be visualized using the original mesh, even though the plane was computed from the point cloud.

---

Interactive 3D Viewer Controls:

• Use mouse to rotate/zoom the object
• Press s to save current camera view
• Press r to reset view
• Press 1 to show/hide point cloud
• Press 2 to show/hide features
• Press 3 to add plane
• Press q to quit

---

Citation

If you find this useful or build on this work, please cite the paper below:

@InProceedings{Aguirre_2026_WACV,
  author    = {Aguirre, Isaac and Sipiran, Ivan},
  title     = {Enhanced Back-Projection of Vision Features for 3D Symmetry Detection},
  booktitle = {Proceedings of the Winter Conference on Applications of Computer Vision (WACV)},
  month     = {-},
  year      = {2026},
  pages     = {-}
}