Quality-Aware Loss (QAL)

WACV 2025 | Paper (arXiv:2511.17824)

Official PyTorch implementation of Quality-Aware Loss for 3D Point Cloud Completion.

Visual Preview

Point-Cloud Completion.

QAL recovers thin structures while controlling spurious points.

Qualitative Comparisons.

Side-by-side comparisons highlighting recall–precision balance versus CD/EMD.

How QAL Works

QAL combines a coverage-weighted nearest-neighbor term with a ground-truth attraction loss, enabling explicit recall–precision control compared with Chamfer/EMD.

Status

🚧 Code release in progress – Expected: [April 2026]

QAL is a drop-in replacement for Chamfer Distance that improves coverage by +4.3 pts on average while recovering thin structures and under-represented regions.

Star/Watch this repo for updates!

Citation

If you find this work helpful, please consider citing:

@article{meshram2025qal,
  title={QAL: A Loss for Recall--Precision Balance in 3D Reconstruction},
  author={Meshram, Pranay and Turkar, Yash and Singh, Kartikeya and Masilamani, Praveen Raj and Adhivarahan, Charuvahan and Dantu, Karthik},
  journal={arXiv preprint arXiv:2511.17824},
  year={2025}
}

Note: This will be updated to the WACV proceedings citation upon publication.

Contact: [pranaywa@buffalo.edu]