How to save predicted video output?

[dlin95123]

When I use a jpg file as the source of predict, I can find a file in the runs folder and it has the image plus the bounding boxes and scores. However, if I use a mp4 file as the source, the file generated in the runs folder is an avi file of size 0. As an example, the following CLI command would generate the output image file:

yolo detect predict model=yolov8s.pt source="c:\users\dlin1\downloads\street.jpg"

However, if I change the source to a mp4 file, I don't get a playable output file. How do I fix that? Thanks.

[zZz-xg]

from ultralytics import YOLO model = YOLO("yolov8s.pt") # pretrained YOLOv8n model results = model.predict(source="c:\users\dlin1\downloads\street.jpg", # or: c:/users/dlin1/downloads/street.jpg ,# street.mp4 Or other format show=True, device=0, # GPU save=True, save_txt=True ) # predict on an image

[glenn-jocher]

@zZz-xg to save the predicted video output when using the yolo detect predict command, ensure that you have the correct dependencies installed and that your environment is properly set up. If you're encountering an issue where the output .avi file is of size 0, this could be due to various reasons such as codec issues, file permission problems, or an interruption during the video processing.

Make sure that:

• Your environment has the necessary codecs to write video files.
• The destination folder has write permissions.
• The source video is not corrupted and is in a readable format.

If the issue persists, you can try running the command with increased verbosity to get more detailed output logs, which might help in diagnosing the problem. Additionally, check the documentation at https://docs.ultralytics.com for any updates or common troubleshooting tips related to video output.

Remember, the show=True flag is used to display the results in a window, but it is not necessary for saving the output video. The command should automatically save the processed video in the runs directory without the need for additional flags. If you continue to face issues, please provide detailed error logs or messages, and we can further investigate the problem.

[XiaMinqi]

@zZz-xg to save the predicted video output when using the yolo detect predict command, ensure that you have the correct dependencies installed and that your environment is properly set up. If you're encountering an issue where the output .avi file is of size 0, this could be due to various reasons such as codec issues, file permission problems, or an interruption during the video processing.

Make sure that:

• Your environment has the necessary codecs to write video files.
• The destination folder has write permissions.
• The source video is not corrupted and is in a readable format.

If the issue persists, you can try running the command with increased verbosity to get more detailed output logs, which might help in diagnosing the problem. Additionally, check the documentation at https://docs.ultralytics.com for any updates or common troubleshooting tips related to video output.

Remember, the show=True flag is used to display the results in a window, but it is not necessary for saving the output video. The command should automatically save the processed video in the runs directory without the need for additional flags. If you continue to face issues, please provide detailed error logs or messages, and we can further investigate the problem.

@glenn-jocher Hi, I'm wondering if it's possible to change the output format as mp4?

[glenn-jocher]

Currently, YOLOv8 saves video outputs in .avi format by default. To convert it to .mp4, you can use a tool like FFmpeg after the prediction process.

[XiaMinqi]

Currently, YOLOv8 saves video outputs in .avi format by default. To convert it to .mp4, you can use a tool like FFmpeg after the prediction process.

Got it! Thanks a a lot!

[glenn-jocher]

You're welcome! If you have any more questions, feel free to ask.

[mariam-162]

You can also change the format to mp4 using openCV..

def convert_avi_to_mp4(input_path, output_path):
    """Convert an AVI video file to MP4 format using OpenCV."""
    cap = cv2.VideoCapture(input_path)
    
    # Get the video's properties (frame width, height, and FPS)
    fps = cap.get(cv2.CAP_PROP_FPS)
    width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))

    # Define the codec and create a VideoWriter object for MP4 output
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # Codec for MP4 format
    out = cv2.VideoWriter(output_path, fourcc, fps, (width, height))

    # Read and write each frame to the new video file
    while True:
        ret, frame = cap.read()
        if not ret:
            break
        out.write(frame)
    
    # Release resources
    cap.release()
    out.release()
    print(f"Video saved as {output_path}")

[glenn-jocher]

@mariam-162 to save the predicted video output in a playable format, ensure the save argument is set to True in your command. If the output is still an AVI file, you can convert it to MP4 using OpenCV as shown in the provided code snippet.

[XiaMinqi]

Thanks for sharing! This is exactly the approach I'm currently using.
However, as using OpenCV leads to more lines in code, I hope if an additional parameter could be set in this tool to control the output format in future version, then mp4 file could be generated gracefully.

[glenn-jocher]

Thank you for your feedback! We'll consider adding an option for output format control in future updates.