Computer Vision News - May 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! Novel method for energy-efficient deep neural networks: Can CNN’s voracious need for data, energy and storage be overcome? Scientists from Oak Ridge National Laboratory propose a new method for the development of energy- efficient deep neural networks capable of solving complex science problems. The proposed technique can turn out very practical for medical applications . Read Now... Autopilotautomaticallymergeintoanother laneinconstructionzone: One of the challenges to Tesla’s Autopilot is detecting static objects in the road, like cones that you typically find in construction areas. Well, the thing is really learning! See by yourself how it reacts to cones now, with no human intervention. Click on the image to watch the video. Understanding Deep Learning through neuron deletion: It’s another great blog by DeepMind, investigating how important are small groups of neurons in deep neural networks and whether more easily interpretable neurons are more important to the network’s computation. The post relates the findings from an experiment in which the network was damaged by deleting individual neurons as well as groups of neurons. Read the post or Read the paper! How and why people think, act, and feel the way they do: Let’s start with a test so easy and funny that you have to try it. It applies IBM’s Watson on a text (that you supply) written by a person whose personality you're interested in. It provides personality insights which are surprising to say the least. No, I won’t tell you what it said about me... Try it here! 48 Computer Vision News Spotlight News Open Letter from leading AI researchers in 30 countries : This one is quite serious. Believe it or not, the South-Korean prestigious institution KAIST is opening a lab in collaboration with major arms company Hanwha Systems to " develop AI technologies to be applied to military weapons, joining the global competition to develop autonomous arms "! Among the top AI researchers calling KAIST to abandon the development of killer robots , we are happy to find Yoshua Bengio and Jürgen Schmidhuber . Read Now...