Extract and sample frames from video streams using OpenCV.
npx skills add https://github.com/bettercallfan/deerflow --skill frame-sampling
Please help me install this Agent Skill. Skill: frame-sampling Source: https://github.com/bettercallfan/deerflow/tree/main/skills/custom/video_surveillance/frame-sampling Command: npx skills add https://github.com/bettercallfan/deerflow --skill frame-sampling
This Skill automates the process of extracting and sampling frames from video streams, saving time and effort in video analysis tasks.
To sample frames from a video file 'input.mp4', use the frame-sampling skill with the following command:
python frame-sampling/scripts/run.py --video input.mp4 --camera-id CAM_DEERFLOW_001 --capture-seconds 10 --interval-seconds 1 --output-dir ./output --config config.json --output frames.json
High-intent search queries and answers about installing and using this skill.
You can extract frames from video streams by providing the raw segment URI and camera ID to this Skill. It uses OpenCV to process the video and applies your chosen sampling strategy to generate stable frame IDs and Frame schema records.
For video surveillance and content analysis, supported frame sampling strategies include interval, fps, keyframe, and adaptive sampling. You select the desired strategy when configuring the extraction parameters to control how frames are captured from the stream.
To generate a frame schema from a video file, run the extraction script with the video path, camera ID, and capture seconds. The Skill processes the video and outputs structured JSON records containing stable frame IDs and metadata for each sampled frame.
No external dependencies are required by the Skill itself to sample keyframes from a video segment. However, since it utilizes OpenCV for video processing, your environment must have OpenCV available to execute the frame extraction scripts successfully.
Required inputs to start sampling frames include camera_id, raw_segment_uri, source_type, sampling_strategy, and capture_seconds. You can also provide optional frame records to generate a complete frame schema for your video surveillance workflow.
zrongAnalyze video content via frames or streams using vision-language models.
bettercallfanTracks detected objects across video frames and outputs Track schema with stable IDs, trajectories, movement states.
bettercallfanDetect objects in video frames using the YOLO algorithm.
buda-aiDetect scene changes and extract keyframes with metadata from video files.