vision-agent 0.2.165__py3-none-any.whl → 0.2.166__py3-none-any.whl

vision_agent/utils/video.py

@@ -11,6 +11,9 @@ import numpy as np
 _LOGGER = logging.getLogger(__name__)
 # The maximum length of the clip to extract frames from, in seconds
 
+_DEFAULT_VIDEO_FPS = 24
+_DEFAULT_INPUT_FPS = 1.0
+
 
 def play_video(video_base64: str) -> None:
     """Play a video file"""
@@ -51,7 +54,9 @@ def _resize_frame(frame: np.ndarray) -> np.ndarray:
 
 
 def video_writer(
-    frames: List[np.ndarray], fps: float = 1.0, filename: Optional[str] = None
+    frames: List[np.ndarray],
+    fps: float = _DEFAULT_INPUT_FPS,
+    filename: Optional[str] = None,
 ) -> str:
     if filename is None:
         filename = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4").name
@@ -78,7 +83,7 @@ def video_writer(
 
 
 def frames_to_bytes(
-    frames: List[np.ndarray], fps: float = 1.0, file_ext: str = ".mp4"
+    frames: List[np.ndarray], fps: float = _DEFAULT_INPUT_FPS, file_ext: str = ".mp4"
 ) -> bytes:
     r"""Convert a list of frames to a video file encoded into a byte string.
 
@@ -101,7 +106,7 @@ def frames_to_bytes(
 # same file name and the time savings are very large.
 @lru_cache(maxsize=8)
 def extract_frames_from_video(
-    video_uri: str, fps: float = 1.0
+    video_uri: str, fps: float = _DEFAULT_INPUT_FPS
 ) -> List[Tuple[np.ndarray, float]]:
     """Extract frames from a video along with the timestamp in seconds.
 
@@ -118,6 +123,16 @@ def extract_frames_from_video(
 
     cap = cv2.VideoCapture(video_uri)
     orig_fps = cap.get(cv2.CAP_PROP_FPS)
+    if not orig_fps or orig_fps <= 0:
+        _LOGGER.warning(
+            f"Input video, {video_uri}, has no fps, using the default value {_DEFAULT_VIDEO_FPS}"
+        )
+        orig_fps = _DEFAULT_VIDEO_FPS
+    if not fps or fps <= 0:
+        _LOGGER.warning(
+            f"Input fps, {fps}, is illegal, using the default value: {_DEFAULT_INPUT_FPS}"
+        )
+        fps = _DEFAULT_INPUT_FPS
     orig_frame_time = 1 / orig_fps
     targ_frame_time = 1 / fps
     frames: List[Tuple[np.ndarray, float]] = []
@@ -129,10 +144,15 @@ def extract_frames_from_video(
             break
 
         elapsed_time += orig_frame_time
+        # This is to prevent float point precision loss issue, which can cause
+        # the elapsed time to be slightly less than the target frame time, which
+        # causes the last frame to be skipped
+        elapsed_time = round(elapsed_time, 8)
         if elapsed_time >= targ_frame_time:
             frames.append((cv2.cvtColor(frame, cv2.COLOR_BGR2RGB), i / orig_fps))
             elapsed_time -= targ_frame_time
 
         i += 1
     cap.release()
+    _LOGGER.info(f"Extracted {len(frames)} frames from {video_uri}")
     return frames

{vision_agent-0.2.165.dist-info → vision_agent-0.2.166.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vision-agent
-Version: 0.2.165
+Version: 0.2.166
 Summary: Toolset for Vision Agent
 Author: Landing AI
 Author-email: dev@landing.ai

{vision_agent-0.2.165.dist-info → vision_agent-0.2.166.dist-info}/RECORD

@@ -28,8 +28,8 @@ vision_agent/utils/execute.py,sha256=FqSOr5gtBeKB1g2hbV6-bhox6qItDQNn2o9efq1w6f4
 vision_agent/utils/image_utils.py,sha256=rm9GfXvD4JrjnqKrP_f2gfq4SzmqYC0IdC1kKwdn6sk,11303
 vision_agent/utils/sim.py,sha256=ZuSS07TUXFGjipmiQoY8TKRmSes7XXCdtU9PI8PC1sw,5609
 vision_agent/utils/type_defs.py,sha256=BE12s3JNQy36QvauXHjwyeffVh5enfcvd4vTzSwvEZI,1384
-vision_agent/utils/video.py,sha256=xbMEoRk13l4fHeQlbvMQhLCn8RNndYmsDhUf01TUeR8,4781
-vision_agent-0.2.165.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-vision_agent-0.2.165.dist-info/METADATA,sha256=jvrYb4IyKp79Sqrhyul6pu0EtEZRewumAZCVR6qWZWg,18034
-vision_agent-0.2.165.dist-info/WHEEL,sha256=7Z8_27uaHI_UZAc4Uox4PpBhQ9Y5_modZXWMxtUi4NU,88
-vision_agent-0.2.165.dist-info/RECORD,,
+vision_agent/utils/video.py,sha256=fOPR48-SuwMbE5eB5rc2F7lVo6k1mVHn26eEJ0QCslc,5602
+vision_agent-0.2.166.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+vision_agent-0.2.166.dist-info/METADATA,sha256=e15d4yNaAJvLCViaBUFo_RNHII88W-y9WgJauEFfbyU,18034
+vision_agent-0.2.166.dist-info/WHEEL,sha256=7Z8_27uaHI_UZAc4Uox4PpBhQ9Y5_modZXWMxtUi4NU,88
+vision_agent-0.2.166.dist-info/RECORD,,