Here I created a web version of the Deep Image Prior algorithm (Original project page, Paper) that runs fully client-side using TensorFlow.JS. It can be used for many purposes such as denoising, inpainting and super-resolution (not implemented yet). Here are some example results for inpainting:
This is an example of how to use NetPrints inside Unity3D for visual programming. It replicates part of the roll-a-ball tutorial where we make a ball roll with user input. For further information on how the example works please visit the link. The completed project is contained in this repository.
Below is an incomplete list of some of my projects
- NetPrints: Visual programming language that can use and output .NET binaries and C# source code
- Cambrian AR: Using machine learning and augmented reality to preview new flooring and painting
- Semantic Relation Classification: Python 3 package for semantic relation extraction and classification using convolutional neural networks
- Keras Neural Painter: Keras-based python scripts for procedural generation of images and seamless videos using random neural networks
- MetaVAE: Meta-learning by applying MAML to an inner variational auto-encoder to automatically learn generative models with few examples. Also contains a more general meta-learning framework and examples to use for classification.