[Task]: Detection of video streams

[EricXuenew]

What is the motivation for this task?

I would like to achieve the function to detect from video streams. Now I can only put pictures into the Anomalib.

Describe the solution you'd like

How to realize the detection of video streams, such as connecting a camera for real-time detection.

Additional context

No response

[Samet Akcay]

@EricXuenew, are you interested to know how this is done, or do you want anomalib to support this by default?

[EricXuenew]

@EricXuenew, are you interested to know how this is done, or do you want anomalib to support this by default?

Thanks for your reply! I would like to know if connecting to a camera is now possible with anomalib?
Also I tried to put in a video, because I see that the code has the option to support video_PATH, but I input a video and get an error!

[EricXuenew]

I would like to implement these features in anomalib v0.7.0, is it feasible, looking forward to your reply!

[Alex Radwan]

You can split frames from the video and do inference

[EricXuenew]

You can split frames from the video and do inference

Like pre-processing it? Does Anomalib have custom algorithm to accomplish it? For example, changing some parameter in the padim.yaml.

[Samet Akcay]

@EricXuenew, you might need to write your own script to detect from frames. It would be something like this -- though you might need do double check the code as it just a pseudo-code

def main():
    # Initialize the webcam (use 0 as the device index to select the default webcam)
    cap = cv2.VideoCapture(0)

    # Check if the webcam is opened correctly
    if not cap.isOpened():
        print("Error: Could not open webcam.")
        return

    # TODO:  Instantiate your inferencer - OpenVINO or Torch Inferencer from anomalib.
    # inferencer = OpenVINOInferencer(path=args.weights, metadata=args.metadata, device=args.device)
    # inferencer = TorchInferencer(path=args.weights, device=args.device)

    while True:
        # Capture frame-by-frame
        ret, frame = cap.read()
        if not ret:
            print("Error: Can't receive frame (stream end?). Exiting ...")
            break

        # Make a prediction using the inferencer
        prediction = inferencer.predict(frame)
        # Alternatively, you could also use `Engine.predict()` method as well. 

        # Display the prediction result on the frame (customize as needed)
        cv2.putText(frame, str(prediction), (50, 50), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0), 2, cv2.LINE_AA)

        # Display the resulting frame
        cv2.imshow('Frame', frame)

        # Break the loop when 'q' is pressed
        if cv2.waitKey(1) & 0xFF == ord('q'):
            break

    # When everything is done, release the capture
    cap.release()
    cv2.destroyAllWindows()

if __name__ == "__main__":
    main()

You could tweak it however you want to. I'm moving this to the Q&A in discussions sections. Feel free to continue there.