Skip to content

holmdk/Video-Prediction-using-PyTorch

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Video-Prediction-using-PyTorch

Alt Text Repository for frame prediction on the MovingMNIST dataset using seq2seq ConvLSTM following either of these guides:

TDS
Github pages

Libraries

Make sure you have the following libraries installed!

python=3.6.8
torch=1.1.0
torchvision=0.3.0
pytorch-lightning=0.7.1
matplotlib=3.1.3
tensorboard=1.15.0a20190708

Getting started

  1. Install the above libraries

  2. Clone this repo

git clone https://github.com/holmdk/Video-Prediction-using-PyTorch.git
cd ./Video-Prediction-using-PyTorch
  1. Run main.py
python main.py
  1. Navigate to http://localhost:6006/ for visualizing results

Results

The first row displays our predictions, the second row the ground truth and the third row the absolute error on a pixel-level. The first 8 columns are the input, followed by output in the final 8 columns. This matches the output from the Tensorboard logging.

After some iterations, we notice that our model is actually generating images of all zeros! This is a common issue people using ConvLSTM reports, however, do not be discouraged! Simply keep training the model, and you should start to see actual and plausible future predictions.

Initial results (500 steps)

Initial

After half an epoch (2500 steps)

Now, we are actually starting to see actual predictions, however blurry they might be. halfepoch

Todo:

  • Add video of predictions by model
  • Implement other video prediction methods (feel free to contribute!)
    • SVG
    • PredRNN+
    • E3D
    • MIM