Learning menu disabled

[Gabriel-Bercaru]

Hello, I cloned this repository and launched nova.exe. I've imported a video and manually annotated a small part of it (at the beginning) with different discrete label (5 predefined labels used).

The goal is to use CML to automatically produce an annotation for the rest of the video, however the LEARNING menu appears gray and I cannot expand, in order to choose Complete Current Annotation (as described in the docs: https://rawgit.com/hcmlab/nova/master/docs/index.html#annotation-completion).

Can you please give some pointers on what might go wrong? Thank you!

[tobiasbaur]

Hello, in order to run the CML step you need to run NOVA in database mode (https://rawgit.com/hcmlab/nova/master/docs/index.html#database)

Best, Tobias

[Gabriel-Bercaru]

Hi Thanks for the previous answer, I have been able to perform the database setup and enabled the CML menu. I could automatically annotate an .mp4 file using the emovoice chain available under bin/cml/chains/video. However I would also like to extract features from the video frames and use them with a neural classifier (such as VGG or RESNET-50) for emotion classification from video.

Is this possible? If so, are there any sample chain files which implement the usage of a neural classifier? An example of a trainer template file would also be helpful (I've seen that the documentation lacks this section - https://rawgit.com/hcmlab/nova/master/docs/index.html#cl-trainer-templates). Thank you!