FeatureFlow-App

This Project is based on FeatureFlow and have been ported to Windows 10. This port inlcudes a user interface which makes it easier to use rather than using command-line.

As of right now, it currently supports videos with a resolution of 640x360 and 1280x720. However as for video with the resolution of 1280x720, this application will automatically split the video into 4 sections and stitch it back all together after each sections has been interpolated.

To Do List

• Allow for videos that are 1920x1080
• Implement proSr or TecoGan
• Use pyscenedectect for scence changes. This will prevent the application from interpolating between scene changes

Bugs

• Does not show about menu
• Does not terminate application when cancelling interpolation dialog box
• Closed Repo

Git Command to pull submodule

git submodule add https://github.com/RNubla/TecoGAN
git submodule update --recursive --remote

PreReq for Building File !!(WARNING: If when building, dont use conda to install packages. Use pip for torchvision and pytorch)

conda create -n open-mmlab python=3.7 -y
conda activate open-mmlab
conda install cudatoolkit=10.0 -y
pip install tensorflow-gpu==1.15.0
pip install ipython==7.4.0
conda install -c conda-forge matplotlib
pip install keras==2.1.2
pip install https://download.pytorch.org/whl/cu100/torch-1.1.0-cp37-cp37m-win_amd64.whl
# pip install https://download.pytorch.org/whl/cu100/torchvision-0.3.0-cp37-cp37m-win_amd64.whl #pytorch 1.1.0
pip install torchvision==0.2.2.post3
conda install -c groakat x264 ffmpeg=4.0.2 -c conda-forge -y
cd mmdetection
git checkout tags/v1.0rc1
pip install -r requirements.txt     # windows use pycocotools-windows
pip install mmcv
pip install -v -e .  # or "python setup.py develop"
pip install scikit-image visdom tqdm prefetch-generator wxPython

BUILD

pyinstaller feflow_ui.py --distpath D:\Programming\Python\FeatureFlow-Build -n FeatureFlow-App -y --clean --add-data .\checkpoints\FeFlow.ckpt;.\checkpoints\ --add-data .\final-model\bdcn_pretrained_on_bsds500.pth;.\final-model\ --add-data .\formbuilder\noname.xrc;.\formbuilder\ --add-binary ffmpeg.exe;.



pyinstaller feflow_ui.py --distpath D:\Programming\Python\FeatureFlow-Build -n FeatureFlow-TecoGAN-App -y --clean --add-data .\checkpoints\FeFlow.ckpt;.\checkpoints\ --add-data .\final-model\bdcn_pretrained_on_bsds500.pth;.\final-model\ --add-data .\formbuilder\noname.xrc;.\formbuilder\ --add-binary ffmpeg.exe;. --add-data .\final-model\TecoGAN.data-00000-of-00001;.\final-model\ --add-data .\final-model\TecoGAN.index;.\final-model\ --add-data .\final-model\TecoGAN.meta\;.\final-model\ --additional-hooks-dir=hooks