model_card.json

{
  "model_name": "roadwork_v3_focal",
  "description": "Vision Transformer (ViT) model for roadwork detection. Trained with validator-matching augmentations, focal loss, label smoothing, and confidence margin loss.",
  "version": "1.0.0",
  "submitted_by": "unknown",
  "submission_time": 1771555076,
  "performance": {
    "accuracy": 0.9839293798098687,
    "mcc": 0.9443160344532321,
    "f1": 0.9902806297056811,
    "auc": 0.9949946578796627,
    "subnet_reward": 0.8961834812863713
  },
  "training": {
    "base_model": "google/vit-base-patch16-224-in21k",
    "training_samples": 34968,
    "validation_samples": 4418,
    "validator_augmentations": true,
    "label_smoothing": 0.05,
    "confidence_margin": 0.1
  }
}