Arty CNN baseline
Multi-head ResNet-50 classifier: genre, style, and artist on a WikiArt subset (pdjota/artyset).
Evaluation (test split)
| Metric | Value |
|---|---|
| Genre (top-1) | 79.61% |
| Style (top-1) | 80.68% |
| Artist (top-1) | 84.50% |
| Artist (top-5) | 97.42% |
- Checkpoint:
best.pt - Arch:
cnn - Epoch (from checkpoint): 8
- Test images: 2438
Limitations
Not for production attribution or forensic ID; academic / demo use.
Dataset used to train pdjota/cnn-baseline
Space using pdjota/cnn-baseline 1
Evaluation results
- Genre accuracy (top-1) on pdjota/artysetLocal eval (scripts/eval_cnn.py)0.796
- Style accuracy (top-1) on pdjota/artysetLocal eval (scripts/eval_cnn.py)0.807
- Artist accuracy (top-1) on pdjota/artysetLocal eval (scripts/eval_cnn.py)0.845
- Artist accuracy (top-5) on pdjota/artysetLocal eval (scripts/eval_cnn.py)0.974