MIT Researchers Made AI That Can Capture Silent Speech

By using cuDNN-accelerated TensorFlow deep learning framework, the researchers trained their model on 31 hours of silently spoken text, which designed to identify subvocalized words from neuromuscular signals.

After that, the neural network tested on 15 people and achieved an accuracy level of 92%.

Real Time 3D Hair Dye Simulation

AR Tech startup ModiFace revealed their deep learning-based live video hair tracking and hair color simulation. They trained their neural networks with 220,000 hair images to make their networks detect hair in each video frame and adjust the coloration of hair in a photorealistic way.

You can try out this new tech at “Hair Color” app on App Store.