metadata
license: mit
tags:
- cad
- brep
- machining-feature-recognition
- graph-neural-network
BrepMFR Model Weights
Model checkpoints for BrepMFR — machining feature recognition in B-rep models through deep learning and domain adaptation.
Files
| Path | Description | F1 Score |
|---|---|---|
trial1_ss/best-epoch41-f1_0.9476.ckpt |
Best overall (full dataset) | 0.9476 |
trial1_ss/last.ckpt |
Last epoch, full dataset | — |
trial1_ss1500/best-epoch46-f1_0.9455.ckpt |
Best, 1500 sample subset | 0.9455 |
trial1_ss1500/last.ckpt |
Last epoch, ss1500 | — |
trial1_ss1000/best-epoch28-f1_0.9146.ckpt |
Best, 1000 sample subset | 0.9146 |
trial1_finetune_ss/best-epoch26-f1_0.9356.ckpt |
Best finetuned | 0.9356 |
pipe_classifier/best-epoch64-f1_0.5517.ckpt |
Pipe classifier (different task) | 0.5517 |
face_seg_heavy/best-epoch43-f1_0.8626.ckpt |
Face segmentation (heavy arch) | 0.8626 |
brepsimplifier/f1_0.9415.ckpt |
BrepFormer for defeaturing | 0.9415 |