What if the robot can recognize objects in 3D?

CSAIL (Computer Science and Artificial Intelligence Laboratory), an MIT computer and artificial intelligence research institute, is studying visual roadmap of objects from visual information using DON (Dense Object Nets) neural network. It takes about 20 minutes to build a visual roadmap. Select the point of the object while viewing the object from all angles. Based on this, 3D objects are recognized.

When a robot learns a thing and then a person talks about a certain point, the robot takes care of the rest. Through this experiment, the robot succeeded in lifting the first shoes and dolls. In the future, it is expected that the robotic arm will be able to recognize and process the object without human help.

Research on computer vision has been around for decades. However, it is a new attempt that neural network learns objects in 3D. Getting things is simple, but it’s one of the more complicated things you’ll have to know about robots. If this research continues, one day, robots can be able to recognize things in real time like humans.

The team will be announcing the results of the project at the Joint Robot Conference. For more information, please click here .