Arwenxu/tmcparts

TMC-Parts

TMC-Parts is a processed multimodal dataset for machining feature recognition in turning-milling combined parts. It accompanies the manuscript TM-MFNet: Structure-Prior-Guided Multimodal Fusion Network for Machining Feature Recognition in Turning-Milling Combined Parts.

Dataset composition

The release contains 34,134 parts with fixed train, validation, and test splits.

Split	Samples
Train	27,307
Validation	3,413
Test	3,414

Each sample provides:

• AAG topology with face and edge attributes;
• UV face grids and edge grids;
• 12 rendered views;
• face-level semantic labels for seven machining categories;
• face-level instance identifiers.

Repository structure

tmcparts/
├── index.json
├── checksums.sha256
├── feature_stats.json
├── splits/
│   ├── train.csv
│   ├── val.csv
│   └── test.csv
└── shards/
    ├── train-0000.sqlite3
    ├── val-0000.sqlite3
    └── ...

The SQLite shards store compressed NumPy arrays and PNG-encoded views. Field definitions are documented in DATA_FORMAT.md.

Download

hf download Arwenxu/tmcparts --repo-type dataset --local-dir TMC-Parts

The dataset can also be downloaded from the Files and versions tab.

Related links

• TMC-Parts project page
• TM-MFNet implementation

License

The processed dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). Original PRT and STEP models are not redistributed.

Citation

Please cite the TMC-Parts dataset and the accompanying TM-MFNet manuscript. Publication and DOI information will be added after publication.