endoreg-db 0.8.6.4__py3-none-any.whl

endoreg_db/models/media/video/video_file_ai.py

@@ -0,0 +1,443 @@
+import logging
+from pathlib import Path
+from typing import TYPE_CHECKING, Optional, List, Dict, Tuple
+from collections import defaultdict, Counter
+
+from ...utils import TEST_RUN as GLOBAL_TEST_RUN, TEST_RUN_FRAME_NUMBER as GLOBAL_N_TEST_FRAMES
+from ...metadata import ModelMeta, VideoPredictionMeta
+
+if TYPE_CHECKING:
+    from .video_file import VideoFile
+    from ...medical.hardware import EndoscopyProcessor
+
+logger = logging.getLogger(__name__)
+
+
+def _extract_text_from_video_frames(
+    video: "VideoFile", frame_fraction: float = 0.001, cap: int = 15
+) -> Optional[Dict[str, str]]:
+    """
+    Extracts text from a sample of video frames using OCR based on processor ROIs.
+    Requires frames to be extracted. Raises ValueError on pre-condition failure.
+    Returns dictionary of extracted text or None if no text found.
+
+    State Transitions:
+        - Pre-condition: Requires state.frames_extracted=True.
+        - Post-condition: No state changes.
+    """
+    from endoreg_db.utils.ocr import (
+        extract_text_from_rois,
+    )  # Local import for dependency isolation
+
+    state = video.get_or_create_state() # Use State helper
+    # --- Pre-condition Check ---
+    if not state.frames_extracted:
+        # Raise exception
+        raise ValueError(f"Frames not extracted for video {video.uuid}. Cannot extract text.")
+    # --- End Pre-condition Check ---
+
+    processor: Optional["EndoscopyProcessor"] = video.processor
+    if not processor:
+        # Raise exception
+        raise ValueError(f"Processor not set for video {video.uuid}. Cannot extract text.")
+
+    try:
+        frame_paths = video.get_frame_paths() # Use Frame helper
+    except Exception as e:
+        logger.error("Error getting frame paths for video %s: %s", video.uuid, e, exc_info=True)
+        raise RuntimeError(f"Could not get frame paths for video {video.uuid}") from e
+
+    n_frames = len(frame_paths)
+    if n_frames == 0:
+        logger.warning("No frame paths found for video %s during text extraction.", video.uuid)
+        return None # Return None if no frames, not an error condition for this function
+
+    # Determine number of frames to process
+    n_frames_to_process = max(1, int(frame_fraction * n_frames))
+    n_frames_to_process = min(n_frames_to_process, cap, n_frames)
+
+    logger.info(
+        "Processing %d frames (out of %d) for text extraction from video %s.",
+        n_frames_to_process,
+        n_frames,
+        video.uuid,
+    )
+
+    # Select evenly spaced frames
+    step = max(1, n_frames // n_frames_to_process)
+    selected_frame_paths = frame_paths[::step][:n_frames_to_process]
+
+    # Extract text from ROIs for selected frames
+    rois_texts = defaultdict(list)
+    errors_encountered = False
+    for frame_path in selected_frame_paths:
+        try:
+            extracted_texts = extract_text_from_rois(frame_path, processor)
+            for roi, text in extracted_texts.items():
+                if text:  # Only append non-empty text
+                    rois_texts[roi].append(text)
+        except Exception as e:
+            # Log error but continue processing other frames
+            logger.error(
+                "Error extracting text from frame %s for video %s: %s", frame_path, video.uuid, e, exc_info=True
+            )
+            errors_encountered = True # Flag that an error occurred
+
+    # Determine the most frequent text for each ROI
+    most_frequent_texts = {}
+    for roi, texts in rois_texts.items():
+        if not texts:
+            most_frequent_texts[roi] = None
+            continue
+        try:
+            counter = Counter(texts)
+            most_common = counter.most_common(1)
+            if most_common:
+                most_frequent_texts[roi] = most_common[0][0]
+            else:
+                most_frequent_texts[roi] = None
+        except Exception as e:
+            logger.error(
+                "Error finding most common text for ROI %s: %s", roi, e, exc_info=True
+            )
+            most_frequent_texts[roi] = None
+
+    if errors_encountered:
+        logger.warning("Errors occurred during text extraction for some frames of video %s. Results may be incomplete.", video.uuid)
+
+    if not most_frequent_texts:
+         logger.info("No text extracted for any ROI for video %s.", video.uuid)
+         return None # Return None if no text found
+
+    logger.info("Extracted text for video %s: %s", video.uuid, most_frequent_texts)
+    return most_frequent_texts
+
+
+def _predict_video_pipeline(
+    video: "VideoFile",
+    model_meta: "ModelMeta",
+    dataset_name: str = "inference_dataset",
+    smooth_window_size_s: int = 1,
+    binarize_threshold: float = 0.5,
+    test_run: bool = False,
+    n_test_frames: int = 10,
+) -> Dict[str, List[Tuple[int, int]]]: # Changed return type to non-optional
+    """
+    Executes the video prediction pipeline using an AI model.
+    Requires frames to be extracted. Raises exceptions on failure.
+
+    State Transitions:
+        - Pre-condition: Requires state.frames_extracted=True.
+        - Post-condition: No state changes directly. (Calling pipeline sets flags).
+    """
+    # Import heavy dependencies locally
+    from ...administration.ai import AiModel
+    try:
+        from ....utils.ai import InferenceDataset, MultiLabelClassificationNet, Classifier
+
+        from ....utils.ai.postprocess import (
+            concat_pred_dicts,
+            make_smooth_preds,
+            find_true_pred_sequences,
+        )
+    except ImportError as e:
+        logger.error("Failed to import endo_ai components: %s. Prediction unavailable.", e, exc_info=True)
+        # Raise exception
+        raise ImportError("Failed to import required AI components for prediction.") from e
+
+    if not test_run and GLOBAL_TEST_RUN:
+        test_run = True
+        n_test_frames = GLOBAL_N_TEST_FRAMES
+        logger.info("Using global TEST_RUN settings for prediction pipeline.")
+
+    state = video.get_or_create_state() # Use State helper
+    # --- Pre-condition Check ---
+    if not state.frames_extracted:
+        # Raise exception
+        raise ValueError(f"Frames not extracted for video {video.uuid}. Prediction aborted.")
+    # --- End Pre-condition Check ---
+
+    # Frame directory check
+    frame_dir = video.get_frame_dir_path() # Use IO helper
+    if not frame_dir or not frame_dir.exists() or not any(frame_dir.iterdir()):
+        # Raise exception
+        raise FileNotFoundError(f"Frame directory {frame_dir} is empty or does not exist for video {video.uuid}. Prediction aborted.")
+
+    model: Optional[AiModel] = model_meta.model
+    if not model:
+        # Raise exception
+        raise ValueError(f"Model not found in ModelMeta {model_meta.name} (Version: {model_meta.version}) for video {video.uuid}. Prediction aborted.")
+
+    # Ensure weights file exists
+    try:
+        weights_path = Path(model_meta.weights.path)
+        if not weights_path.exists():
+            # Raise exception
+            raise FileNotFoundError(f"Model weights file {weights_path} not found for {model_meta.name} (Video: {video.uuid}). Prediction aborted.")
+    except Exception as e:
+        logger.error("Error accessing model weights path for %s (Video: %s): %s", model_meta.name, video.uuid, e, exc_info=True)
+        raise RuntimeError(f"Error accessing model weights for {model_meta.name}") from e
+
+    # Get or create VideoPredictionMeta
+    try:
+        _video_prediction_meta, created = VideoPredictionMeta.objects.get_or_create(
+            video_file=video, model_meta=model_meta
+        )
+        if created:
+            logger.info(
+                "Created new VideoPredictionMeta for video %s, model %s.",
+                video.uuid,
+                model_meta.name,
+            )
+        else:
+            logger.info(
+                "Found existing VideoPredictionMeta for video %s, model %s.",
+                video.uuid,
+                model_meta.name,
+            )
+        # video_prediction_meta.save() # Save is handled by get_or_create
+    except Exception as e:
+        logger.error(
+            "Failed to get or create VideoPredictionMeta for video %s, model %s: %s", video.uuid, model_meta.name, e, exc_info=True
+        )
+        # Raise exception
+        raise RuntimeError("Failed to get or create VideoPredictionMeta") from e
+
+    # --- Dataset Preparation ---
+    datasets = {
+        "inference_dataset": InferenceDataset,
+        # Add other dataset types here if needed
+    }
+    dataset_model_class = datasets.get(dataset_name)
+    if not dataset_model_class:
+        # Raise exception
+        raise ValueError(f"Dataset class '{dataset_name}' not found for video {video.uuid}. Prediction aborted.")
+
+    try:
+        paths = video.get_frame_paths() # Use Frame helper
+        if not paths:
+            raise FileNotFoundError(f"No frame paths returned by get_frame_paths for {frame_dir} (Video: {video.uuid})")
+    except Exception as e:
+        logger.error(
+            "Error listing or getting frame files from %s for video %s: %s", frame_dir, video.uuid, e, exc_info=True
+        )
+        raise RuntimeError(f"Error getting frame paths from {frame_dir}") from e
+
+    logger.info("Found %d frame files in %s for video %s.", len(paths), frame_dir, video.uuid)
+
+    crop_template = video.get_crop_template() # Use Meta helper
+    string_paths = [p.as_posix() for p in paths]
+    crops = [crop_template] * len(paths) # Assuming same crop for all frames
+
+    if test_run:
+        logger.info("TEST RUN: Using first %d frames for video %s.", n_test_frames, video.uuid)
+        string_paths = string_paths[:n_test_frames]
+        crops = crops[:n_test_frames]
+        if not string_paths:
+             # Raise exception
+            raise ValueError(f"Not enough frames ({len(paths)}) for test run (required {n_test_frames}) for video {video.uuid}.")
+
+    try:
+        ds_config = model_meta.get_inference_dataset_config()
+        ds = dataset_model_class(string_paths, crops, config=ds_config)
+        logger.info("Created dataset '%s' with %d items for video %s.", dataset_name, len(ds), video.uuid)
+        if len(ds) > 0:
+            sample = ds[0] # Get a sample for debugging shape
+            logger.debug("Sample shape: %s", sample.shape)
+    except Exception as e:
+        logger.error(
+            "Failed to create dataset '%s' for video %s: %s", dataset_name, video.uuid, e, exc_info=True
+        )
+        # Raise exception
+        raise RuntimeError(f"Failed to create dataset '{dataset_name}'") from e
+
+    # --- Model Loading ---
+    try:
+        # Check if CUDA is available
+        import torch
+        if torch.cuda.is_available():
+            try:
+                # Try loading on GPU first
+                ai_model_instance = MultiLabelClassificationNet.load_from_checkpoint(
+                    checkpoint_path=weights_path.as_posix(), # Ensure path is string
+                )
+                # Attempt to move to GPU
+                _ = ai_model_instance.cuda()
+                logger.info("Loaded model on GPU for video %s.", video.uuid)
+            except RuntimeError as cuda_err:
+                # If GPU loading fails, fall back to CPU
+                logger.warning("GPU loading failed for video %s: %s. Falling back to CPU.", video.uuid, cuda_err)
+                ai_model_instance = MultiLabelClassificationNet.load_from_checkpoint(
+                    checkpoint_path=weights_path.as_posix(),
+                    map_location='cpu'
+                )
+                logger.info("Loaded model on CPU for video %s.", video.uuid)
+        else:
+            # No CUDA available, load directly on CPU
+            logger.info("CUDA not available. Loading model on CPU for video %s.", video.uuid)
+            ai_model_instance = MultiLabelClassificationNet.load_from_checkpoint(
+                checkpoint_path=weights_path.as_posix(),
+                map_location='cpu'
+            )
+
+        _ = ai_model_instance.eval() # Set to evaluation mode
+        classifier = Classifier(ai_model_instance, verbose=True) # Assuming Classifier exists
+        logger.info("AI model loaded successfully for video %s from %s.", video.uuid, weights_path)
+    except Exception as e:
+        logger.error(
+            "Failed to load AI model for video %s from %s: %s", video.uuid, weights_path, e, exc_info=True
+        )
+        # Raise exception
+        raise RuntimeError(f"Failed to load AI model from {weights_path}") from e
+
+    # --- Inference ---
+    try:
+        logger.info("Starting inference on %d frames for video %s...", len(string_paths), video.uuid)
+        predictions = classifier.pipe(string_paths, crops)
+        logger.info("Inference completed for video %s.", video.uuid)
+    except Exception as e:
+        logger.error("Inference failed for video %s: %s", video.uuid, e, exc_info=True)
+        # CUDA-OOM Fallback: Speicher freigeben und CPU versuchen
+        try:
+            import torch, gc
+            is_oom = isinstance(e, (getattr(torch.cuda, 'OutOfMemoryError', RuntimeError), RuntimeError)) and (
+                'out of memory' in str(e).lower() or 'cuda out of memory' in str(e).lower()
+            )
+        except Exception:
+            is_oom = False
+        if 'torch' in globals() or 'torch' in locals():
+            try:
+                import torch  # ensure available in this scope
+                if torch.cuda.is_available() and is_oom:
+                    logger.warning("CUDA OOM detected. Freeing CUDA cache and retrying on CPU…")
+                    try:
+                        torch.cuda.empty_cache()
+                        gc.collect()
+                    except Exception:
+                        pass
+                    try:
+                        # Move model to CPU and retry inference
+                        _ = ai_model_instance.cpu()
+                        classifier = Classifier(ai_model_instance, verbose=True)
+                        predictions = classifier.pipe(string_paths, crops)
+                        logger.info("Inference completed on CPU after CUDA OOM for video %s.", video.uuid)
+                    except Exception as e2:
+                        logger.error("CPU fallback inference failed for video %s: %s", video.uuid, e2, exc_info=True)
+                        # Raise exception
+                        raise RuntimeError("Inference failed") from e2
+                else:
+                    # Raise exception
+                    raise RuntimeError("Inference failed") from e
+            except Exception:
+                # Raise exception
+                raise RuntimeError("Inference failed") from e
+        else:
+            # Raise exception
+            raise RuntimeError("Inference failed") from e
+
+    # --- Post-processing ---
+    try:
+        logger.info("Post-processing predictions for video %s...", video.uuid)
+        readable_predictions = [classifier.readable(p) for p in predictions]
+
+        merged_predictions = concat_pred_dicts(readable_predictions)
+
+        fps = video.get_fps() # Use Meta helper
+        if not fps:
+            logger.warning(
+                "Video FPS is unknown for %s. Smoothing/sequence calculations might be inaccurate. Using default 30 FPS.",
+                video.uuid,
+            )
+            fps = 30 # Default FPS if unknown
+
+        smooth_merged_predictions = {}
+        for key in merged_predictions.keys():
+            smooth_merged_predictions[key] = make_smooth_preds(
+                prediction_array=merged_predictions[key],
+                window_size_s=smooth_window_size_s,
+                fps=fps,
+            )
+
+        binary_smooth_merged_predictions = {}
+        for key in smooth_merged_predictions.keys():
+            binary_smooth_merged_predictions[key] = (
+                smooth_merged_predictions[key] > binarize_threshold
+            )
+
+        sequences = {}
+        for label, prediction_array in binary_smooth_merged_predictions.items():
+            sequences[label] = find_true_pred_sequences(prediction_array)
+
+        logger.info(
+            "Post-processing completed for video %s. Found sequences for labels: %s",
+            video.uuid,
+            list(sequences.keys()),
+        )
+        return sequences if sequences is not None else {}
+
+    except Exception as e:
+        logger.error("Post-processing failed for video %s: %s", video.uuid, e, exc_info=True)
+        # Raise exception
+        raise RuntimeError("Post-processing failed") from e
+
+
+def _predict_video_entry(
+    video: "VideoFile",
+    model_name: str,
+    model_meta_version: Optional[int] = None,
+    dataset_name: str = "inference_dataset",
+    smooth_window_size_s: int = 1,
+    binarize_threshold: float = 0.5,
+    test_run: bool = GLOBAL_TEST_RUN,
+    n_test_frames: int = GLOBAL_N_TEST_FRAMES,
+    save_results: bool = True, # Note: save_results is handled in video_file.py now
+):
+    """Entry point called from VideoFile.predict_video. Imports and calls the main prediction logic."""
+    from endoreg_db.models import AiModel, ModelMeta # Local import
+
+    try:
+        ai_model = AiModel.objects.get(name=model_name)
+        if not model_meta_version:
+            model_meta = ai_model.get_latest_version()
+            logger.info("Using latest ModelMeta version %s for model %s.", model_meta_version, model_name)
+        else:
+            model_meta = ai_model.get_version(model_meta_version)
+            logger.info("Using specified ModelMeta version %s for model %s.", model_meta_version, model_name)
+
+        logger.info("Using ModelMeta: %s (Version: %s)", model_meta.name, model_meta.version)
+    except ModelMeta.DoesNotExist:
+        logger.error("ModelMeta '%s' (Version: %s) not found.", model_name, model_meta_version)
+        raise
+
+    # --- Explicitly pass only the arguments expected by _predict_video_pipeline ---
+    predicted_sequences = _predict_video_pipeline(
+        video=video,
+        model_meta=model_meta, # Pass the fetched ModelMeta object
+        dataset_name=dataset_name,
+        smooth_window_size_s=smooth_window_size_s,
+        binarize_threshold=binarize_threshold,
+        test_run=test_run,
+        n_test_frames=n_test_frames,
+    )
+    # --- End Explicit Arguments ---
+
+    # Return the sequences and the ModelMeta object used
+    return predicted_sequences, model_meta
+
+
+def _extract_text_information(
+    video: "VideoFile", frame_fraction: float = 0.001, cap: int = 15
+) -> Optional[Dict[str, str]]:
+    """Facade function to call the text extraction logic."""
+    logger.info("Attempting text extraction for video %s.", video.uuid)
+
+    extracted_data = _extract_text_from_video_frames(
+        video=video, frame_fraction=frame_fraction, cap=cap
+    )
+
+    if extracted_data is not None:
+        logger.info("Text extraction successful for video %s.", video.uuid)
+    else:
+        logger.warning("Text extraction returned no data for video %s.", video.uuid)
+
+    return extracted_data