Google travaille sur une technologie de reconnaissance d'objets capable de tout détecter

A lot of companies are now focusing on introducing enhanced automatic object recognition system in order to provide users a whole new level of experience. In a recent article de blog published on Google Blog, Google explained some of the new techniques that it will be using for the object recognition mechanism that will help to detect almost everything present in a living room.

Google said that it is team of researchers scored the highest points in few different categories at the annual ImageNet computer vision competition held last month. Google’s team took the lead in the classification and detection categories. Google’s team openly presented their ideas with everyone present at the competition in order to bolster the progress in the field by forming respective community with fellow researchers.

The three categories that were part of the competition are classification, classification with localization and detection. The first category lays down an algorithm that is used to assign the correct labels to an image and things present in it. The classification with localization is used to measure the efficiency of how correctly the images are labeled and how accurately the objects are underlined accordingly. The last category, detection, is similar to the second one except that it focuses more on the final evaluation criteria of the labels and its outlining.

The research team included two members who are also a part of PhD program at Google and they are actively working on this concept with the rest of the team members. This new technological advancement will help general public recognize objects in an image easily. It is said that as soon as the new object recognition research is finalized, it will be directly available to Google products including photo search, image search, YouTube and self driving cars etc.