{
  "name": "FER2025",
  "description": "FER2025 is a large-scale, balanced facial emotion dataset containing 1,589,810 images across 7 emotion classes (Angry, Disgust, Fear, Happy, Neutral, Sad, Surprise). It is designed for feature extraction, model training, and benchmarking in computer vision and deep learning.",
  "version": "1.0.0",
  "total_images": 1589810,
  "classes": ["Angry", "Disgust", "Fear", "Happy", "Neutral", "Sad", "Surprise"],
  "formats": ["jpg", "jpeg", "png"],
  "recommended_usage": [
    "Feature extraction with CNNs or ViT",
    "Training deep learning models",
    "Benchmarking facial expression recognition algorithms"
  ],
  "dataset_structure": [
    "Sharded TAR/WebDataset: FER2025/ClassName.tar (images + .cls labels)"
  ],
  "license": "Research-use-only",
  "usage_notes": "FER2025 is for research and non-commercial purposes only. Users must respect the licenses of the original datasets included (FER2013, AffectNet, CK+, RAF-DB, etc.) and provide proper attribution.",
  "source_datasets": [
    {"name": "FER2013", "license": "CC BY-NC"},
    {"name": "AffectNet", "license": "Research-use-only"},
    {"name": "CK+", "license": "Research-use-only"},
    {"name": "RAF-DB", "license": "Research-use-only"}
  ],
  "citation": "Adhavan M. FER2025: Large-Scale Balanced Facial Expression Dataset. 2025. URL: https://huggingface.co/datasets/imadhavan/FER2025"
}