Document all models + link dedicated repos for the trained ones (RMIB, MeshSeg)
Browse files
README.md
CHANGED
|
@@ -34,4 +34,45 @@ CC0-1.0), trained on CC0 content from ambientCG / Poly Haven. Converted with
|
|
| 34 |
|
| 35 |
QtMeshEditor downloads these on first use into `<AppData>/ai_models/pbr/`.
|
| 36 |
|
| 37 |
-
More info in [QtMesh Cloud website](https://qtmesh.dev)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
| 35 |
QtMeshEditor downloads these on first use into `<AppData>/ai_models/pbr/`.
|
| 36 |
|
| 37 |
+
More info in [QtMesh Cloud website](https://qtmesh.dev)
|
| 38 |
+
## Texture upscaling
|
| 39 |
+
|
| 40 |
+
`RealESRGAN_x2plus.onnx`, `RealESRGAN_x4plus.onnx` — 2×/4× super-resolution.
|
| 41 |
+
ONNX re-exports of **Real-ESRGAN** ([xinntao](https://github.com/xinntao/Real-ESRGAN),
|
| 42 |
+
**BSD-3-Clause**). Downloaded into `<AppData>/ai_models/pbr/`. Credit: xinntao.
|
| 43 |
+
|
| 44 |
+
## Auto-rig skeleton prediction (UniRig)
|
| 45 |
+
|
| 46 |
+
`unirig/encoder.onnx`, `unirig/decoder.onnx`, `unirig/embed.onnx` — ML skeleton
|
| 47 |
+
prediction for unrigged meshes. These are **ONNX re-exports** of
|
| 48 |
+
**[VAST-AI/UniRig](https://huggingface.co/VAST-AI/UniRig)** (SIGGRAPH 2025 — MIT
|
| 49 |
+
code + MIT weights, trained on Articulation-XL2.0 / CC-BY-4.0). Converted with
|
| 50 |
+
`scripts/export-unirig-onnx.py` in the QtMeshEditor repo. Downloaded into
|
| 51 |
+
`<AppData>/ai_models/unirig/`. Credit for the weights: VAST-AI-Research.
|
| 52 |
+
|
| 53 |
+
## Animation in-betweening (RMIB) — trained by us
|
| 54 |
+
|
| 55 |
+
`inbetween/rmib.onnx` — fills the gap between two keyframes with smooth
|
| 56 |
+
intermediate motion. **Trained from scratch** by the QtMeshEditor project on the
|
| 57 |
+
permissive [CMU Graphics Lab Motion Capture Database](http://mocap.cs.cmu.edu).
|
| 58 |
+
Beats spherical-linear interpolation by >2× on held-out CMU motion.
|
| 59 |
+
**Dedicated repo:** [fernandotonon/QtMeshEditor-rmib-inbetween](https://huggingface.co/fernandotonon/QtMeshEditor-rmib-inbetween).
|
| 60 |
+
Downloaded into `<AppData>/ai_models/inbetween/`. License: CC-BY-4.0.
|
| 61 |
+
|
| 62 |
+
## Mesh part segmentation — trained by us
|
| 63 |
+
|
| 64 |
+
`segment/meshseg.onnx` — predicts head / torso / arm / leg labels per point
|
| 65 |
+
(PointNet++-style). **Trained from scratch** by the QtMeshEditor project on
|
| 66 |
+
**synthetic, permissively-derived** data (per-vertex labels from rigged-humanoid
|
| 67 |
+
bone weights — CC0). Sidesteps the non-commercial ShapeNet-Part / PartNet
|
| 68 |
+
datasets. **Dedicated repo:** [fernandotonon/QtMeshEditor-mesh-segmentation](https://huggingface.co/fernandotonon/QtMeshEditor-mesh-segmentation).
|
| 69 |
+
Downloaded into `<AppData>/ai_models/segment/`. License: CC-BY-4.0.
|
| 70 |
+
|
| 71 |
+
---
|
| 72 |
+
|
| 73 |
+
These models power the AI-assisted authoring features in
|
| 74 |
+
**[QtMeshEditor](https://github.com/fernandotonon/QtMeshEditor)** and its
|
| 75 |
+
companion **QtMesh Cloud** ([qtmesh.dev](https://qtmesh.dev)). Each downloads on
|
| 76 |
+
first use and runs locally (offline). Mixed licenses per model as noted above
|
| 77 |
+
(CC0 / BSD-3 / MIT-derived / CC-BY-4.0); see each section + the QtMeshEditor
|
| 78 |
+
`THIRD_PARTY_AI_MODELS.md`.
|