Skip to content

Latest commit

 

History

History
35 lines (26 loc) · 860 Bytes

File metadata and controls

35 lines (26 loc) · 860 Bytes

FlowColorization

Dataset has to be donwloaded separately.

Contents:

dataset_preparation.ipynb: NB for generating the proper training data from DAVIS-videos: RGB (and grey) images, and optical flows

UNet_train_noFlow.ipynb: NB for training the UNet Model

Demo.ipynb: Render Colored video using a trained Model.

ToDo's:

  • Data:
    • test/train split
    • data samples including >1 frame difference
  • Network:
    • UNet, VAE, GAN ?
    • think about Loss function
  • Results
    • show performance of fully trained Model (make NB)

Layout I&O: NN IO Design

Requirements (so far):

  • opencv-python
  • scikit-image
  • pytorch
  • torchvision
  • matplotlib