Hi.
I have read your description and understood.
I have many experiences in deep learning, especially object classification, object detection, and generation modeling. Also, I am familiar with computer vision.
You want to create a dataset for video generation using DVD-GAN?
For this, I will use the object classification or object detection CNN model to infer the topics of the videos and find the frames that contain the interested objects. Next, I'll use openCV( and ffmpeg library) to assemble them into 5secs part of each video. The script language will be be chosen of what you want. Of course, if it is Python is, Ok.
I expect to chat with you for this project.
Regards.