Dataset Viewer
The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
DS5 — Final-Training Dataset
Status: READY TO POPULATE (post-implementation)
| Aspect | Value |
|---|---|
| Purpose | Primary deliverable — ML/CD-SEM benchmarking dataset |
| Size | 100,000 samples @1024×1024 |
| Image | 1024×1024, 16-bit, 1.0 nm/px |
| Master seed | 5005 |
| Expected volume | ~400 GB |
| Generation config | generation_configs/ds5_final_training.yml |
| Validation gates | L1, L2, L3, L4 (sampling), L5 |
| Release criteria | All gates pass; benchmark-readiness review; stakeholder sign-off on structure weighting |
Structure Weighting (pending stakeholder review)
| Structure | Weight | Count |
|---|---|---|
| dense_ls | 20.0% | 20,000 |
| contact | 15.0% | 15,000 |
| via | 10.0% | 10,000 |
| iso_line | 15.0% | 15,000 |
| fin | 10.0% | 10,000 |
| gate | 10.0% | 10,000 |
| trench | 8.0% | 8,000 |
| sti | 5.0% | 5,000 |
| bimaterial | 4.0% | 4,000 |
| pitch_std | 3.0% | 3,000 |
Splits (frozen, deterministic)
| Split | Count | Fraction |
|---|---|---|
| Train | 70,000 | 70% |
| Validation | 15,000 | 15% |
| Test | 15,000 | 15% |
Stratified by structure type; disjoint sample sets (no leakage).
Contents (populated by generator)
ds5_final_training/
├── dataset_index.json
├── dataset_schema.txt
├── LICENSE (CC BY 4.0)
├── README.md
├── images/*.tiff
├── ground_truth/*.json
├── metadata/*_config.json, *_metadata.json
├── splits/train.txt, val.txt, test.txt
└── logs/
Generation Command (post-implementation)
semicon-sim batch --config ../generation_configs/ds5_final_training.yml
semicon-sim --validate ds5_final_training --level L5
No images exist yet — this directory is specified, not populated.
- Downloads last month
- 53