OTVision 0.6.5__tar.gz → 0.6.7__tar.gz

{otvision-0.6.5 → otvision-0.6.7}/.github/workflows/release-testpypi.yml

@@ -26,14 +26,37 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
+
+      - name: Fetch all tags
+        run: git fetch --tags --unshallow
+
       - name: Set up Python
         uses: actions/setup-python@v5
         with:
           python-version: "3.12"
+
+      - name: Get latest tag and create new version
+        id: get_version
+        run: |
+          # Get the latest tag in format v*.*.* (ignoring tags like "nightly")
+          LATEST_TAG="$(git tag -l "v*.*.*" | sort -V | tail -n 1 || echo 'v0.0.0')"
+          # Extract version without 'v' prefix
+          VERSION="${LATEST_TAG#v}"
+          # Split version into parts
+          IFS='.' read -r MAJOR MINOR PATCH <<< "$VERSION"
+          # Increment patch version
+          PATCH="$((PATCH + 1))"
+          # Create new version with incremented patch
+          INCREMENTED_VERSION="$MAJOR.$MINOR.$PATCH"
+          DATE_TIME="$(date +'%Y%m%d%H%M%S')"
+          NEW_VERSION="${INCREMENTED_VERSION}a${DATE_TIME}"
+          echo "Final version: $NEW_VERSION"
+          echo "new_version=$NEW_VERSION" >> "$GITHUB_OUTPUT"
+
       - name: Update Python Package Version
         uses: platomo/update-version-py-action@v1.0.1
         with:
-          version: nightly
+          version: ${{ steps.get_version.outputs.new_version }}
           file-path: "OTVision"
       - name: Install pypa/build
         shell: bash

{otvision-0.6.5 → otvision-0.6.7}/.pre-commit-config.yaml

@@ -63,14 +63,13 @@ repos:
           - pre-commit==3.8.0
           - pytest-cov==5.0.0
           - pytest==8.3.3
-          - torch==2.3.1
           - torchvision==0.18.1
           - twine==6.1.0
           - types-PyYAML
           - types-flake8
           - types-geopandas
           - types-jsonschema
-          - types-setuptools
+          - types-torch
           - types-tqdm
           - types-ujson
           - ultralytics==8.3.94

{otvision-0.6.5 → otvision-0.6.7}/OTVision/application/detect/detected_frame_factory.py

@@ -20,6 +20,7 @@ class DetectedFrameFactory:
 
         return DetectedFrame(
             source=frame[FrameKeys.source],
+            output=frame[FrameKeys.output],
             no=frame[FrameKeys.frame],
             occurrence=frame[FrameKeys.occurrence],
             detections=detections,

{otvision-0.6.5 → otvision-0.6.7}/OTVision/detect/otdet_file_writer.py

@@ -62,7 +62,7 @@ class OtdetFileWriter:
         detect_config = config.detect
 
         actual_frames = len(event.frames)
-        if (expected_duration := detect_config.expected_duration) is not None:
+        if expected_duration := detect_config.expected_duration:
             actual_fps = actual_frames / expected_duration.total_seconds()
         else:
             actual_fps = actual_frames / source_metadata.duration.total_seconds()

{otvision-0.6.5 → otvision-0.6.7}/OTVision/detect/rtsp_input_source.py

@@ -27,10 +27,12 @@ from OTVision.domain.time import DatetimeProvider
 
 RTSP_URL = "rtsp://127.0.0.1:8554/test"
 RETRY_SECONDS = 1
+DEFAULT_READ_FAIL_THRESHOLD = 5
 
 
 class Counter:
     def __init__(self, start_value: int = 0) -> None:
+        self._start_value = start_value
         self.__counter = start_value
 
     def increment(self) -> None:
@@ -39,6 +41,9 @@ class Counter:
     def get(self) -> int:
         return self.__counter
 
+    def reset(self) -> None:
+        self.__counter = self._start_value
+
 
 class RtspInputSource(InputSourceDetect):
 
@@ -68,12 +73,17 @@ class RtspInputSource(InputSourceDetect):
     def flush_buffer_size(self) -> int:
         return self.stream_config.flush_buffer_size
 
+    @property
+    def fps(self) -> float:
+        return self.config.convert.output_fps
+
     def __init__(
         self,
         subject: Subject[FlushEvent],
         datetime_provider: DatetimeProvider,
         frame_counter: Counter,
         get_current_config: GetCurrentConfig,
+        read_fail_threshold: int = DEFAULT_READ_FAIL_THRESHOLD,
     ) -> None:
         super().__init__(subject)
         self._datetime_provider = datetime_provider
@@ -82,6 +92,11 @@ class RtspInputSource(InputSourceDetect):
         self._get_current_config = get_current_config
         self._current_stream: str | None = None
         self._current_video_capture: VideoCapture | None = None
+        self._stream_start_time: datetime = self._datetime_provider.provide()
+        self._current_video_start_time = self._stream_start_time
+        self._outdated = True
+        self._read_fail_threshold = read_fail_threshold
+        self._consecutive_read_fails = 0
 
     @property
     def _video_capture(self) -> VideoCapture:
@@ -106,20 +121,30 @@ class RtspInputSource(InputSourceDetect):
         return self._current_video_capture
 
     def produce(self) -> Generator[Frame, None, None]:
-        start_time = self._datetime_provider.provide()
+        self._stream_start_time = self._datetime_provider.provide()
+        self._current_video_start_time = self._stream_start_time
         while not self.should_stop():
             if (frame := self._read_next_frame()) is not None:
                 self._frame_counter.increment()
+                occurrence = self._datetime_provider.provide()
+
+                if self._outdated:
+                    self._current_video_start_time = occurrence
+                    self._outdated = False
 
                 yield Frame(
                     data=convert_frame_to_rgb(frame),  # YOLO expects RGB
                     frame=self.current_frame_number,
                     source=self.rtsp_url,
-                    occurrence=self._datetime_provider.provide(),
+                    output=self.create_output(),
+                    occurrence=occurrence,
                 )
                 if self.flush_condition_met():
-                    self._notify(start_time)
-        self._notify(start_time)
+                    self._notify()
+                    self._outdated = True
+                    self._frame_counter.reset()
+
+        self._notify()
 
     def _init_video_capture(self, source: str) -> VideoCapture:
         cap = VideoCapture(source)
@@ -136,10 +161,22 @@ class RtspInputSource(InputSourceDetect):
     def _read_next_frame(self) -> ndarray | None:
         successful, frame = self._video_capture.read()
         if successful:
+            self._consecutive_read_fails = 0
             return frame
+        self._consecutive_read_fails += 1
+
+        if self._consecutive_read_fails >= self._read_fail_threshold:
+            self._try_reconnecting_stream()
+
         logger().debug("Failed to grab frame")
         return None
 
+    def _try_reconnecting_stream(self) -> None:
+        self._video_capture.release()
+        self._current_video_capture = None
+        if not self.should_stop() and self._current_stream is not None:
+            self._current_video_capture = self._init_video_capture(self._current_stream)
+
     def should_stop(self) -> bool:
         return self._stop_capture
 
@@ -152,24 +189,16 @@ class RtspInputSource(InputSourceDetect):
     def flush_condition_met(self) -> bool:
         return self.current_frame_number % self.flush_buffer_size == 0
 
-    def _notify(self, start_time: datetime) -> None:
+    def _notify(self) -> None:
         frame_width = int(self._video_capture.get(CAP_PROP_FRAME_WIDTH))
         frame_height = int(self._video_capture.get(CAP_PROP_FRAME_HEIGHT))
-        fps = self.config.convert.output_fps
-        _start_time = calculate_start_time(
-            start_time, self.current_frame_number, fps, self.flush_buffer_size
-        )
         frames = (
             self.flush_buffer_size
             if self.current_frame_number % self.flush_buffer_size == 0
             else self.current_frame_number % self.flush_buffer_size
         )
-        duration = timedelta(seconds=round(frames / fps))
-        output_filename = (
-            f"{self.stream_config.name}_FR{round(fps)}"
-            f"_{_start_time.strftime(DATETIME_FORMAT)}.mp4"
-        )
-        output = str(self.stream_config.save_dir / output_filename)
+        duration = timedelta(seconds=round(frames / self.fps))
+        output = self.create_output()
         self._subject.notify(
             FlushEvent.create(
                 source=self.rtsp_url,
@@ -177,22 +206,17 @@ class RtspInputSource(InputSourceDetect):
                 duration=duration,
                 source_width=frame_width,
                 source_height=frame_height,
-                source_fps=fps,
-                start_time=_start_time,
+                source_fps=self.fps,
+                start_time=self._current_video_start_time,
             )
         )
 
-
-def calculate_start_time(
-    start: datetime, current_frame_number: int, fps: float, flush_buffer_size: int
-) -> datetime:
-    offset_in_frames = (
-        current_frame_number // flush_buffer_size - 1
-    ) * flush_buffer_size
-    if offset_in_frames == 0:
-        return start
-    offset_in_seconds = offset_in_frames / fps
-    return start + timedelta(seconds=offset_in_seconds)
+    def create_output(self) -> str:
+        output_filename = (
+            f"{self.stream_config.name}_FR{round(self.fps)}"
+            f"_{self._current_video_start_time.strftime(DATETIME_FORMAT)}.mp4"
+        )
+        return str(self.stream_config.save_dir / output_filename)
 
 
 def convert_frame_to_rgb(frame: ndarray) -> ndarray:

{otvision-0.6.5 → otvision-0.6.7}/OTVision/detect/timestamper.py

@@ -61,6 +61,7 @@ class VideoTimestamper(Timestamper):
             data=frame[FrameKeys.data],
             frame=frame[FrameKeys.frame],
             source=frame[FrameKeys.source],
+            output=frame[FrameKeys.output],
             occurrence=occurrence,
         )
 

{otvision-0.6.5 → otvision-0.6.7}/OTVision/detect/video_input_source.py

@@ -116,6 +116,7 @@ class VideoSource(InputSourceDetect):
                                     FrameKeys.data: rotated_image,
                                     FrameKeys.frame: frame_number,
                                     FrameKeys.source: str(video_file),
+                                    FrameKeys.output: str(video_file),
                                 }
                             )
                         else:
@@ -124,6 +125,7 @@ class VideoSource(InputSourceDetect):
                                     FrameKeys.data: None,
                                     FrameKeys.frame: frame_number,
                                     FrameKeys.source: str(video_file),
+                                    FrameKeys.output: str(video_file),
                                 }
                             )
                         counter += 1
@@ -206,8 +208,9 @@ class VideoSource(InputSourceDetect):
     def __add_occurrence(self, timestamper: Timestamper, frame: dict) -> Frame:
         updated = timestamper.stamp(frame)
         return Frame(
-            data=updated["data"],
-            frame=updated["frame"],
-            source=updated["source"],
-            occurrence=updated["occurrence"],
+            data=updated[FrameKeys.data],
+            frame=updated[FrameKeys.frame],
+            source=updated[FrameKeys.source],
+            output=updated[FrameKeys.output],
+            occurrence=updated[FrameKeys.occurrence],
         )

{otvision-0.6.5 → otvision-0.6.7}/OTVision/domain/frame.py

@@ -20,6 +20,7 @@ class FrameKeys:
     frame: Literal["frame"] = "frame"
     source: Literal["source"] = "source"
     occurrence: Literal["occurrence"] = "occurrence"
+    output: Literal["output"] = "output"
 
 
 class Frame(TypedDict):
@@ -35,6 +36,7 @@ class Frame(TypedDict):
     data: Optional[ndarray]
     frame: int
     source: str
+    output: str
     occurrence: datetime
 
 
@@ -49,6 +51,7 @@ class DetectedFrame:
         no (FrameNo): Frame number.
         occurrence (datetime): Time stamp, at which frame was recorded.
         source (str): Source from where frame was obtained, e.g. video file path.
+        output (str): Output file name, e.g. video file name.
         detections (Sequence[Detection]): A sequence of Detections occurring in frame.
         image (Optional[ndarray]): Optional image data of frame.
     """
@@ -56,6 +59,7 @@ class DetectedFrame:
     no: FrameNo
     occurrence: datetime
     source: str
+    output: str
     detections: Sequence[Detection]
     image: Optional[ndarray] = None
 
@@ -64,6 +68,7 @@ class DetectedFrame:
             no=self.no,
             occurrence=self.occurrence,
             source=self.source,
+            output=self.output,
             detections=self.detections,
             image=None,
         )
@@ -153,6 +158,7 @@ class TrackedFrame(DetectedFrame):
             no=self.no,
             occurrence=self.occurrence,
             source=self.source,
+            output=self.output,
             finished_tracks=self.finished_tracks,
             detections=detections,
             image=self.image,

{otvision-0.6.5 → otvision-0.6.7}/OTVision/track/parser/chunk_parser_plugins.py

@@ -62,6 +62,7 @@ class JsonChunkParser(ChunkParser):
                 no=int(key) + frame_offset,
                 occurrence=occurrence,
                 source=str(file),
+                output=str(file),
                 detections=detections,
                 image=None,
             )

{otvision-0.6.5 → otvision-0.6.7}/OTVision/track/tracker/tracker_plugin_iou.py

@@ -222,6 +222,7 @@ class IouTracker(Tracker):
             no=frame.no,
             occurrence=frame.occurrence,
             source=frame.source,
+            output=frame.output,
             detections=tracked_detections,
             image=frame.image,
             finished_tracks=set(finished_track_ids),

{otvision-0.6.5 → otvision-0.6.7}/OTVision/version.py

@@ -1,4 +1,4 @@
-__version__ = "v0.6.5"
+__version__ = "v0.6.7"
 
 
 def otdet_version() -> str:

{otvision-0.6.5 → otvision-0.6.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: OTVision
-Version: 0.6.5
+Version: 0.6.7
 Summary: OTVision is a core module of the OpenTrafficCam framework to perform object detection and tracking.
 Project-URL: Homepage, https://opentrafficcam.org/
 Project-URL: Documentation, https://opentrafficcam.org/overview/
@@ -27,7 +27,6 @@ Requires-Dist: numpy==2.1.1; sys_platform != 'win32'
 Requires-Dist: opencv-python==4.10.0.84
 Requires-Dist: pandas==2.2.3
 Requires-Dist: pyyaml==6.0.2
-Requires-Dist: setuptools==74.0.0
 Requires-Dist: torch==2.3.1
 Requires-Dist: torchvision==0.18.1
 Requires-Dist: tqdm==4.67.1

{otvision-0.6.5 → otvision-0.6.7}/requirements.txt

@@ -9,7 +9,6 @@ numpy==1.26.4; sys_platform == 'win32' # ultralytics requires numpy <= 2.1.1; to
 opencv-python==4.10.0.84 # only upgrade when https://github.com/opencv/opencv/issues/27091 is resolved
 pandas==2.2.3
 PyYAML==6.0.2
-setuptools==74.0.0
 # Follow instructions at https://pytorch.org to install torch with CUDA
 torch==2.3.1 #  NOTE: Omit if using torch with CUDA and install manually
 torchvision==0.18.1 # NOTE: Omit if using torch with CUDA and install manually

{otvision-0.6.5 → otvision-0.6.7}/tests/application/detect/test_detected_frame_factory.py

@@ -16,6 +16,7 @@ class TestDetectedFrameFactory:
 
         assert actual == DetectedFrame(
             source=given_frame[FrameKeys.source],
+            output=given_frame[FrameKeys.output],
             no=given_frame[FrameKeys.frame],
             occurrence=given_frame[FrameKeys.occurrence],
             detections=given_detections,
@@ -28,6 +29,7 @@ class TestDetectedFrameFactory:
             frame=1,
             occurrence=datetime(2020, 1, 1, 12, 1, 1),
             source="path/to/source.mp4",
+            output="path/to/output.mp4",
         )
 
     def create_detections(self) -> list[Detection]:

{otvision-0.6.5 → otvision-0.6.7}/tests/application/detect/test_timestamper.py

@@ -98,6 +98,7 @@ def create_frame_without_occurrence(frame_number: int) -> dict:
         FrameKeys.data: Mock(),
         FrameKeys.frame: frame_number,
         FrameKeys.source: SOURCE,
+        FrameKeys.output: SOURCE,
     }
 
 
@@ -112,5 +113,6 @@ def create_expected_frame(raw_data: dict, occurrence: datetime) -> Frame:
         data=raw_data[FrameKeys.data],
         frame=raw_data[FrameKeys.frame],
         source=raw_data[FrameKeys.source],
+        output=raw_data[FrameKeys.output],
         occurrence=occurrence,
     )

{otvision-0.6.5 → otvision-0.6.7}/tests/detect/test_detected_frame_buffer.py

@@ -47,6 +47,7 @@ class TestDetectedFrameBuffer:
         frame_number = 1
         occurrence = datetime(2020, 1, 1, 12, 0, 0)
         source = "my_source"
+        output = "path/to/output.mp4"
         detections: list[Detection] = create_mocks(3)
         image = Mock()
 
@@ -54,6 +55,7 @@ class TestDetectedFrameBuffer:
             no=frame_number,
             occurrence=occurrence,
             source=source,
+            output=output,
             detections=detections,
             image=image,
         )
@@ -64,6 +66,7 @@ class TestDetectedFrameBuffer:
             no=frame_number,
             occurrence=occurrence,
             source=source,
+            output=output,
             detections=detections,
             image=None,
         )

otvision-0.6.5/tests/detect/test_odet.py → otvision-0.6.7/tests/detect/test_otdet.py

@@ -19,6 +19,7 @@ GIVEN_ACTUAL_DURATION = timedelta(
     hours=36, seconds=33, milliseconds=12, microseconds=15
 )
 EXPECTED_VIDEO_LENGTH = "36:00:33"
+DETECTED_FRAME_OUTPUT = "path/to/output.mp4"
 
 
 def create_expected_video_metadata(
@@ -109,6 +110,7 @@ def create_detected_frame(source: str, frame_number: int) -> DetectedFrame:
     )
     return DetectedFrame(
         source=source,
+        output=DETECTED_FRAME_OUTPUT,
         no=frame_number,
         detections=[detection],
         occurrence=datetime(2020, 1, 1, 12, 0, 10),