Browsing Tag
Google AI
Google AI Implements Reinforcement Learning for Chip Design Optimization
Researchers at Google AI have devised a new method of optimizing chips designs for future machine learning requirements and use cases. The method implements reinforcement learning to remodel chip placement for any application specific integrated circuit, to reduce power, performance…
Google AI Releases XTREME Benchmark for Multilingual AI Models
With XTREME, Google aims to incorporate multiple languages for cross-lingual generalization of machine learning models. XTREME specifically caters to linguistically differing and typologically diverse languages of the world. It stands for TRansfer Evaluation of Multilingual Encoders.
Google Achieves Higher Accuracy of Self-Supervised and Semi-Supervised Learning
Google has demonstrated improved self-supervised and semi-supervised learning on ImageNet dataset with SimCLR or A Simple Framework for Contrastive Learning of Visual Representations. The code and pre-trained SimCLR models are available on GitHub.
Google AI Release 3D Reconstruction Tools in Time for CVPR 2020
Google AI has developed and released a computer vision based pattern recognition benchmark leveraging structure-from-motion for 3D reconstruction. The updated methodology features pose information for each of its nearly 25 thousand images, now available to the public under Image Matching…
Google AI Rolls Out WaveNetEQ in Duo for Enhanced Audio Calls
Researchers at Google AI have trained a recurrent neural network (RNN) model to fill in the gaps left by packet data loss in online calls. The WaveNetEQ model is trained on speech generated under 48 languages by over 100 speakers…
Google Adds Art Transfer to its Arts & Culture Platform and App
Google’s new Art Transfer is the bridge between legendary artists like van Gogh, da Vinci or Frida Kahlo, and the current tech-enabled environment! The feature is a part of the Google Arts & Culture mobile app, which utilizes a unique…
Google AI Delivers a Computationally Efficient Reinforcement Learning Agent
Reinforcement learning neural networks are generally time consuming, involve high-costs, and deliver limited rates of training or ‘learning’. Researchers at Google AI have overcome this by centralizing inference on accelerators like GPUs and TPUs to train on millions of frames…
Google AI and DeepMind Release Reinforcement Learning Agent Dreamer
Research carried out under Google AI and DeepMind’s collaboration for model-based reinforcement learning (RL) has resulted in the development and release of the source code of Dreamer, an RL agent that uses behavior learning for action predictions. Apt to its…
Google AI Integrate Motion Sense for Hands-Free Interaction with Pixel 4 & Pixel 4 XL
Extensive collaborative efforts by Google in regard to the development of hardware, software and gesture-driven machine learning models for a machine perceptive highly sensitive (unlike its military equivalents) radar sensor has resulted in the Soli Radar Sensor. The short-range sensor…
Google AI Release Neural Tangents for Infinite Node Deep Neural Networks
Google AI’s new Neural Tangents is an open source library, now available to download, for researchers delving into the exploratory aspects of deep neural networks in regard to infinite number of nodes. This infinite aspect signifies high node density available…
Google AI Facilitate 3D Object Detection on Mobile Devices
Google AI have developed and released an on-device machine learning pipeline on MediaPipe, complemented with a mobile app for 3D object detection in real-time. The MediaPipe Objectron leverages augmented reality and machine learning to determine the object position, orientation and…
