Computer Vision News - June 2019

LeCun - self-supervised learning for images and videos: If you missed Yann LeCun ’s now famous post about "unsupervised“ and " self-supervised learning “ (how could you?), here it is. And if you missed also Yann LeCun’s interview , you should read it now . ReadMore… Google releases AI training data set with 5M images: Let’s start with Google: they open-sourced a database called Google-Landmarks-v2 , which follows a previous one from last year. This database contains over 5 million images of more than 200,000 different landmarks collected from photographers around the world. They also announce two new Kaggle challenges : Landmark Recognition 2019 and Landmark Retrieval 2019. Read… Google’s lung cancer detection outperforms 6 radiologists: This powerful AI uses a patient’s current and prior CT volumes to predict the risk of lung cancer , and it allows to optimize the screening process via computer assistance and automation. RSIP Vision also detects and segments lung tumors ! See page 16 in this mag. Read… 39 Computer Vision News Artificial Intelligence Spotlight News Computer Vision News has found great new stories, written somewhere else by somebody else. We share them with you, adding a short comment. Enjoy! TensorFlow Model Optimization Toolkit - Pruning API: MATLAB 2019a - More AI, Systems-Engineering Support: people are starting to use MATLAB’s deep learning platform. The guys at electronicdesign.com have reviewed the “ quite a few enhancements and additions to an already formidable development package ”. After this intro, you can’t but Read It! Scientists help Artificial Intelligence outsmart hackers: AI needs to outsmart hackers and neutralize hostile adversarial attacks. The alternative, when AI is vulnerable to patterns added by attackers, is to see this threat become commonplace. Here is how researchers at ICLR want to give AI a defensive edge. ReadMore… This neural net would like to deliver these petitions: Talk to Transformer: how a modern neural network completes your text: 3 more links worth clicking: