reliquary-deepbump-normal

DeepBump (color → Normal CNN) ONNX export. Powers Reliquary's in-PBR-slot "DeepBump Direct" generator and the standalone AI Normal sidebar module.

Source

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

• Repo: https://github.com/HugoTini/DeepBump
• Paper: DeepBump — Hugo Tini, 2020 (no associated paper).
• Pretrained weights origin: Upstream deepbump256.onnx (fixed 256×256) hosted in the DeepBump repo's master branch.

License

Inherited from upstream DeepBump (GPL-3.0). Non-commercial / GPL-compatible projects only. Distributing a project that bundles these weights forces the whole project under GPL-3.0 at distribution.

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/deepbump.ts
• Input shape / format: [N, 1, 256, 256] — single-channel grayscale luminance (Rec. 709 luma Y = 0.299R + 0.587G + 0.114B), pixels in [0, 1]. Runtime tile-inferences at this fixed shape.
• Output shape / format: [N, 3, 256, 256] — tangent-space normal map (DirectX convention by default; flip channel 1 sign for OpenGL).

Files

• deepbump_normal.onnx — ONNX graph (25.5 MB)
• 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.