Add 'has_occlusion' parameter with default value True

This commit introduces a new parameter 'has_occlusion' with a default value of True. The tracking mechanism now stops early under two conditions: when more than half of the tracking instances are lost, or when there is no occlusion detected in the video and at least one instance is lost during tracking. In other cases, the system will log a message indicating the event.

annolid/segmentation/cutie_vos/predict.py

@@ -16,7 +16,6 @@
 from hydra import compose, initialize
 from annolid.segmentation.cutie_vos.model.cutie import CUTIE
 from annolid.segmentation.cutie_vos.inference.inference_core import InferenceCore
-from annolid.segmentation.cutie_vos.inference.utils.args_utils import get_dataset_cfg
 from pathlib import Path
 import gdown
 from annolid.utils.devices import get_device
@@ -108,7 +107,8 @@ def process_video_with_mask(self, frame_number=0,
                                 labels_dict=None,
                                 pred_worker=None,
                                 recording=True,
-                                output_video_path=None
+                                output_video_path=None,
+                                has_occlusion=True,
                                 ):
         if mask is not None:
             num_objects = len(np.unique(mask)) - 1
@@ -183,16 +183,18 @@ def process_video_with_mask(self, frame_number=0,
                             message_with_index = message + \
                                 delimiter + str(current_frame_index)
                             logger.info(message)
-                            # If half of the instances are missing, then stop the prediction.
-                            if len(num_instances_in_current_frame) < self.num_tracking_instances / 2:
+                            # Stop the prediction if more than half of the instances are missing,
+                            # or when there is no occlusion in the video and one instance loses tracking.
+                            if (not has_occlusion or
+                                len(num_instances_in_current_frame) < self.num_tracking_instances / 2
+                                ):
+                                pred_worker.stop_signal.emit()
+                                # Release the video capture object
+                                cap.release()
+                                # Release the video writer if recording is set to True
+                                if recording:
+                                    self.video_writer.release()
                                 return message_with_index
-                            pred_worker.stop_signal.emit()
-                            # Release the video capture object
-                            cap.release()
-                            # Release the video writer if recording is set to True
-                            if recording:
-                                self.video_writer.release()
-                            return message_with_index
                         if recording:
                             visualization = overlay_davis(frame, prediction)
                             # Write the frame to the video file