params.yaml

media_processing:
  video_processing:
    input_video_dir: data/raw/original/
    output_video_dir: data/interim/reduced/
    reduction_factor: 4
    mirror_videos: True

  photo_processing:
    input_photo_dir: data/raw/photos/
    output_photo_dir: data/interim/photos/
    mirror_photos: True

labeling:
  labels:
    m: meathook
    n: nutcracker
    l: l-hang
    o: other pose or transition
    r: reverse meathook
    b: back lever
    f: front lever
  skip_seconds: 1
  input_video_dir: data/interim/reduced/
  input_photo_dir: data/interim/photos/
  output_dir: data/interim/labeled/
  force_relabel: 
    - file1.jpg
    - file2.mov
  force_relabel_all: True

features:
  input_video_dir: data/interim/reduced/
  output_video_dir: data/processed/videos/
  save_annotated_video: False
  input_photo_dir: data/interim/photos/
  output_photo_dir: data/processed/photos/
  save_annotated_photo: True
  interim_landmarks_directory: data/interim/landmarks
  interim_features_directory: data/interim/features
  labeled_dir: data/interim/labeled
  final_features_directory: data/processed/features

model_training:
  MLflow_config:
    experiment_name: "XGB_Optimization_and_Training"
    run_names:
      main: "Main_Pipeline_Run"
      hyperparameter_optimization: "Hyperparameter_Optimization"
      final_training: "Retrain_on_Train_Dataset"
      prod_training: "Prod_Model_FULL_Dataset_Train"
  model_type: xgb
  final_features_filepath: data/processed/features/final_features.csv
  test_size: 0.2
  target_column: label
  predictions_dir: data/results/
  optimize_hyperparams: True
  optuna_config:
    study_name: xgb_optimization_study
    delete_existing_study: True
  num_trials: 5
  score_metric: f1_weighted  #f1_micro, f1_weighted, 'recall', 'precision', 'roc_auc', etc. as needed (not f1 for multiclass)
  train_prod_model: True

production_model:
  model_name: production_test
  dest_dir: models/prod/xgb/