to detect anomalies in video data do we have to convert the video to csv file using encoder(which). can anyone help me in this?
I suggest you read through this thread: Introducing new project: image2sdr.
You would need to sample images out of the video to create an SDR stream. Then create a Network with an SP and TM. Then you could get anomaly values.
What type of video are you planning on analyzing?
the video will be an episode containing advertisements. I have to show the anamolies whenever advertisement starts and ends.
We’ve talked about the idea of using DL to extract features from live video still images and feeding the features SDR into HTM for temporal analysis. See: Step by step do do HTM from a image