I have trained a model using my datasets, how can I use the model to detect anomaly in the video (.mp4) ?

[zhanghuayu-seu]

What is the motivation for this task?

Real-time anomaly detection

Describe the solution you'd like

We use lightning_inference.py to achieve anomaly detection of a picture successfully, but find no proper functions to detect directly in the video.

Additional context

No response

[samet-akcay]

@djdameln, can you help here?

[alexriedel1]

You should write a function that reads single frames from the video and pass them to the inference function

[glucasol]

@zhanghuayu-seu Were You able to do that? How much FPS did you got with this real time inference?

[samet-akcay]

@glucasol, if speed is your concern for a video inference, you could potentially utilise OpenVINO's async inference, which significantly improves the FPS for such use cases. For more information, you could refer to this link:
https://docs.openvino.ai/2023.0/notebooks/115-async-api-with-output.html

Let us know if you have any more questions.