cnn-baseline / README.md
pdjota's picture
Update README.md
de32675 verified
metadata
license: apache-2.0
language: en
pipeline_tag: image-classification
library_name: pytorch
datasets:
  - pdjota/artyset
tags:
  - pytorch
  - torchvision
  - resnet-50
  - image-classification
  - multi-task
  - wikiart
  - art
model-index:
  - name: Arty CNN baseline
    results:
      - task:
          type: image-classification
        dataset:
          name: pdjota/artyset
          type: pdjota/artyset
        metrics:
          - name: Genre accuracy (top-1)
            type: accuracy
            value: 0.796144
          - name: Style accuracy (top-1)
            type: accuracy
            value: 0.806809
          - name: Artist accuracy (top-1)
            type: accuracy
            value: 0.844955
          - name: Artist accuracy (top-5)
            type: accuracy
            value: 0.974159
        source:
          name: Local eval (scripts/eval_cnn.py)
          url: https://huggingface.co/docs/hub/model-cards

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.