Code Reference

These pages document the core Python modules that power the AutoML wildfire pipeline.

• Prelabeling Modules
  • YOLO Prelabeling
  • Grounding DINO Prelabeling
  • Matching Logic
• Human-in-the-Loop Module
  • export_versioned_results()
  • import_tasks_to_project()
  • run_human_review()
  • setup_label_studio()
  • transform_reviewed_results_to_labeled()
• Augmentation
  • augment_dataset()
  • augment_images()
  • build_augmentation_transform()
• Training
  • find_latest_model()
  • load_train_config()
  • train_model()
• Distillation
  • build_optimizer_and_scheduler()
  • calculate_gradient_norm()
  • freeze_layers()
  • head_features_decoder()
  • load_checkpoint()
  • load_models()
  • log_training_metrics()
  • prepare_dataset()
  • save_checkpoint()
  • save_final_model()
  • start_distillation()
  • train_epoch()
  • train_loop()
• Quantization
  • fp16_quantization()
  • imx_quantization()
  • onnx_quantization()
  • quantize_model()
• File Cleaning & Archiving
  • clean_pipeline_workspace()
• Utility Scripts
  • create_data_yaml()
  • create_distill_yaml()
  • create_quantize_yaml()
  • detect_device()
  • draw_yolo_bboxes()
  • load_config()
  • prepare_quantization_data()
  • prepare_training_data()