Description of Problem & Solution
The issue with this software is it is still very manual. I have it working. I can run the various parts. It would be more useful if it created an HTML table and file with the clips and times instead of images and a csv.
The key issue is that the scene detection and splitting need to be accurate enough not to have 15 adjustments to the thresholds and parameters. I can easily bring any media file into final cut pro (or MPEG Streamclip). The mark and split. Handbrake, for instance, can do something similar.
I literally have 50,000 30 minutes to 2-hour videos to test against. None of the ones I have tried are even close out of the box. Since this uses OpenCV, it should be trainable to be more and more accurate.
What propose is a means of adding learning to the OpenCV
Media Examples:
I can upload the source video but it is 1.3GB.
Proposed Implementation:
I would propose some means for learning. Probably tied into the GUI for the product.
Alternative Solutions:
Description of Problem & Solution
The issue with this software is it is still very manual. I have it working. I can run the various parts. It would be more useful if it created an HTML table and file with the clips and times instead of images and a csv.
The key issue is that the scene detection and splitting need to be accurate enough not to have 15 adjustments to the thresholds and parameters. I can easily bring any media file into final cut pro (or MPEG Streamclip). The mark and split. Handbrake, for instance, can do something similar.
I literally have 50,000 30 minutes to 2-hour videos to test against. None of the ones I have tried are even close out of the box. Since this uses OpenCV, it should be trainable to be more and more accurate.
What propose is a means of adding learning to the OpenCV
Media Examples:
I can upload the source video but it is 1.3GB.
Proposed Implementation:
I would propose some means for learning. Probably tied into the GUI for the product.
Alternative Solutions: