Some tools to modify video content based on well established Deep Learning algorithms for segmentation, inpainting, 3d modelling...
Still work in progress, but some preliminary results available:
Preliminary results of object removal (original video)
pip .... (TODO)On the root folder:
git clone https://github.com/nestormh/neuralgymneuralgym repository had to be included in this one to be compatible with the latest version of tensorflow. I avoided created a fork since just some parts of the code were upgraded and could be confusing. Original code can be checked here: neuralgym.
Changes:
- Some functions have been renamed / reconfigured for the newer tensorflow version.
- Checkpoints provided were incompatible with the actual code, so some variable names had to be changed.
Models included at the moment are required for image inpainting. You can use these available at:
- Models are directly retrieved from tensorflow repositories.
Check generative_inpainting repository for more information.