The selected demos are visualization created for research progress, which explores the unique possibilities of machine learning to reconfigure and analyze moving images through experimenting with AI models. The research particularly focuses on the analysis of the rhythm of old films and new clip generation.

The video on the left shows a neural network trained to predict future frames of an old B&W movie. This clip shows the original movie on the left and the predictions on the right. The other images are visualizations for latent features embedded in Cross Transformer and image reconstruction using AdaBins Model.

​

Publication 2024 CVPR AI Art:

https://thecvf-art.com/project/rhuthmus/