Open tamal-thetaonelab opened 1 week ago
Thanks for your interest in our approach. From what I understand, you're mainly interested in maintaining only the latest 1 hour of footage without needing to keep the memory of the historical data, right? In this case, you can truncate from the left of the input sequence to maintain a moving 1-hour window for the latest streaming frames.
What can be a good approach for handling streaming media / very long footages ?
What I understood, currently the model can handle videos of an hour length or so. Keeping this restriction, how can I implement a sliding window (an hour, say) on a streaming CCTV footage. Let's say I want queries to be answered always on last 1 hour of the video.
Any advise is greatly appreciated.