# ATM Anomaly Detection Training Configuration

# Model
model:
  base: yolov8n.pt  # Options: yolov8n, yolov8s, yolov8m, yolov8l, yolov8x
  pretrained: true
  classes: 14  # Number of anomaly classes

# Training
training:
  epochs: 100
  batch_size: 16
  image_size: 640
  learning_rate: 0.001
  optimizer: AdamW
  weight_decay: 0.0005
  momentum: 0.937
  
  # Augmentation
  augmentation:
    hsv_h: 0.015
    hsv_s: 0.7
    hsv_v: 0.4
    degrees: 5.0
    translate: 0.1
    scale: 0.5
    shear: 2.0
    perspective: 0.0
    flipud: 0.0
    fliplr: 0.5
    mosaic: 1.0
    mixup: 0.0
    copy_paste: 0.0
  
  # Loss
  box_loss_gain: 7.5
  cls_loss_gain: 0.5
  dfl_loss_gain: 1.5

# Data
data:
  train: data/splits/train
  val: data/splits/val
  test: data/splits/test
  
  # Class names (must match database schema)
  names:
    0: physical_damage
    1: vandalism
    2: graffiti
    3: dirt_debris
    4: obstruction
    5: skimming_device
    6: suspicious_attachment
    7: out_of_service
    8: screen_damage
    9: cash_jam
    10: receipt_jam
    11: lighting_failure
    12: camera_blind
    13: network_down

# Validation
validation:
  conf_threshold: 0.25
  iou_threshold: 0.45
  max_detections: 300
  
# Output
output:
  checkpoint_dir: models/checkpoints
  export_dir: models/exports
  log_dir: logs
  
# Hardware
hardware:
  device: auto  # auto, cpu, cuda:0
  workers: 8
  
# Logging
logging:
  project: atm_anomaly
  name: experiment_001
  save_period: 10
  plot: true