# model_card.yaml

model_name: "AONomaly Detection Model"
model_type: "autoencoder"
language: "en"
license: "mit"

tags:
  - anomaly-detection
  - autoencoder
  - edge-ai
  - openvino
  - onnx
  - computer-vision
  - unsupervised-learning

task_categories:
  - anomaly-detection
  - image-classification

library_name: "pytorch"

datasets:
  - name: "Casting Product Image Dataset"
    source: "https://www.kaggle.com/datasets/ravirajsinh45/real-life-industrial-dataset-of-casting-product"

metrics:
  - name: "Reconstruction Error Threshold"
    type: "MSE"
    value: 0.01

model-index:
  - name: "AONomaly Detection Model"
    results:
      - task:
          type: "anomaly-detection"
          name: "Casting Defect Detection"
        dataset:
          name: "Casting Product Image Dataset"
          type: "image"
        metrics:
          - name: "MSE Reconstruction Error"
            type: "float"
            value: 0.01

inference:
  input_format: "Grayscale image (128x128)"
  output_format: "Reconstructed image + anomaly score"

intended_use:
  primary_use: "Industrial defect inspection via anomaly detection."
  limitations:
    - "Requires consistent lighting and background conditions."
    - "Trained specifically on metal casting images."

author:
  name: "Arunima Surendran"
  role: "AI Developer & Researcher"
  repository: "https://github.com/arunimakanavu/aonmalydetectionmodel"
  email: "N/A"

framework_versions:
  pytorch: "2.2.0"
  openvino: "2024.1"
  onnx: "1.15.0"