Computer Vision News - February 2018

Computer Vision News lists some of the great stories that we have just found somewhere else. We share them with you, adding a short comment. Enjoy! How can A.I. Generate Believable Fake Photos: Sorry, guys: the lady in the photo looks familiar but she does not exist. She was generated by software under development at NVIDIA in no less than 18 days of image processing. This software can analyze thousands of (real) celebrity snapshots, recognize common patterns, and create new images that look much the same - but are still a little different. What will be the effect of this work on the spread of misinformation online, you can only guess. Read Now... Stickers make CV software hallucinate things that aren’t there If you want to fool an AI vision system, this article is for you. Adversarial images (but also audio and text) distorts visual features to exploit weaknesses in the way computers look at the world to make them see stuff that isn’t there. Kind of optical illusions for Computer Vision Software . Read Now... The Last 5 Years In Deep Learning: Oh, no, not another story about Deep Learning ! Well, this is a very interesting sum up of the hype, the impact, the progress and future perspectives of Deep Learning. We like this piece, courtesy of Adit Deshpande . Read it here! 3D Face Reconstruction from a Single Image: Let’s start with something funny that you can try by yourself ! This is a 3D reconstruction of the face of @POTUS. How did we do it? We just used a lovely tool, courtesy of Aaron Jackson , PhD Student at the University of Nottingham . I tried to upload a pic of myself, but I think this one is way cooler. BTW, it takes only a few seconds and you can play with the result as you wish. Try it here! Do you always hear that ‘correlation does not imply causation’? Here is the brilliant proof! Go package for computer vision using OpenCV 3.x and beyond: GoCV 10 Computer Vision News Spotlight News How AI is transforming the criminal justice system: This is far more serious. Should judicial courts use AI systems to speed, streamline or reduce bias in criminal sentencing ? Believe it or not, they already do . But far from being a simple idea, it is actually very intricate, as it involves many severe difficulties: think only at how different the effects of two separate sets of training data would be! Read Now...