Adding new videos without having to label from the start


I am adding new videos to a project that has already been run from A to Z.
When i added the videos i was asked to extract frames for labeling. The model was first trained on a video that is part 1 of a video of 4 parts, and the other parts are just a continuation of part 1. So three more video parts. Could i skip the frame labeling in this case? JFTR the video does not include new unexpected events, it is like a video of a fish in fish tank swimming around for 1 hour divided into 15 mins shorter videos.

Do you need to improve your model and retrain because of poor tracking? If yes, then you need to label extracted frames (see the docs about refining).

If the first video tracked well, and you just want to analyze more video, you don’t need to add those videos to the project. You can just analyze them.

Adding videos to the project means that you are adding them to the training set.


Thank you Brandon for your answer!
How do i just analyze them? I just select the video in the analyze video stage and that is it?

Yes. Just select the videos where ever they are, or select the folder and set the file extension to match your videos (to analyze all videos of that type in that folder. All of the outputs will be saved in the same folder as the videos.


1 Like