Feature Request: Support Video Input for Frame-by-Frame Inference Using 2D Models #115
Description
Hi,
I'm working with live cell imaging data. My typical segmentation workflow involves training a 2D segmentation model on a set of static images, and then applying that trained model to segment cells in videos, frame by frame. I've previously used the ZeroCostDL4Mic Colab notebooks, which support this approach.
I was trying something similar using the BiaPy 2D instance segmentation notebook. I successfully trained a model, but when I tried to use the inference notebook on a video file, it understandably didn't work—since the notebook/configuraion file expects individual image files, not videos.
Would it be possible to add support to the inference notebook for loading a video file directly, processing it frame by frame using a trained 2D model, and then saving the segmented output as a video (multiframe TIFF)? This would make it much easier to work with time-lapse data and batch-process multiple videos, without needing to manually extract frames beforehand.