You can deploy this model and start earning money today!

Wszelkiebro/
reliquary-biglama-inpainting

Name: reliquary-biglama-inpainting
Availability: InStock

Views

No views yet

reliquary-biglama-inpainting

LaMa (Large Mask Inpainting) re-exported to ONNX with dynamic input axes and opset 18 (DFT-aware). Powers Reliquary's AI Auto-Tiling research path.

Source

ONNX re-export of the upstream model — original credits to the authors:

Repo: https://github.com/advimman/lama
Paper: "Resolution-robust Large Mask Inpainting with Fourier Convolutions" — Suvorov et al., WACV 2022
Pretrained weights origin: HuggingFace mirror smartywu/big-lama (official LaMa team upload, ~400 MB PyTorch checkpoint)

License

Inherited from upstream LaMa (Apache 2.0). Free for commercial and non-commercial use with attribution.

Usage in Reliquary

This model is consumed by Reliquary, a browser-based PBR texture editor. The ONNX file is loaded via ONNX Runtime Web in the browser — no Python or server-side inference.

Runtime integration point: src/lib/ai/biglama.ts (TBD)
Input shape / format: [N, 4, H, W] — RGB (channels 0-2) + binary mask (channel 3, 1.0 = inpaint, 0.0 = keep). H/W dynamic.
Output shape / format: [N, 3, H, W] — inpainted RGB, same spatial dims as input.

Files

biglama.onnx — ONNX graph (4.7 MB)
biglama.onnx.data — external weights (194.6 MB, present because the model exceeds the ONNX protobuf size limit)
README.md — this file

Disclaimer

This is a re-export from the original PyTorch checkpoint to a browser-deployable ONNX format with dynamic input axes. The model weights, architecture, and training methodology are entirely the work of the original authors cited above. This repo is a hosting convenience for Reliquary's browser-side ML — it does NOT modify the model.

For the full conversion pipeline (including version-compat shims for PyTorch 2.6+, opset selection, dynamic-axes details), see Reliquary's PYTHON_CONVERSIONS_GUIDE.md.

Wszelkiebro/reliquary-biglama-inpainting