State of Art Video Prediction 2021

This repository references to the paper "A Review on Deep Learning Techniques for Video Prediction", and provides a unified collection of the best-performing architectures in order to explore the advancements in video prediction in 2021.

Architectures

Deterministic

• CrevNet Paper - Code (unofficial)
• E3D-LSTM Paper - Code (unofficial)

Probabilistic

• SVG Paper - Code (unofficial)
• Improved VRNN Paper - Code (unofficial)

Experiments

Fully Attention-Based

I am currently working on the application of fully attention-based models that attempt to overcome the limitations with Convolutional Neural Networks.

• Generative TimeSformer Code (unofficial)
• TimeSformer Paper - Code (unofficial)
• ViViT Paper - Code (unofficial)

Thank you

Credits list to be released soon. All the credits are reserved to the sources mentioned above.