Live Science
–
Illustration of man with robot. Recently MIT researchers developed a robot eye that is more optimal in interpreting objects in front of it. His abilities were even like those of a human eye.
–
Nationalgeographic.co.id—Robot indeed human aids, but he is not perfect, including how his eyes process objects in front of him that are more than one object. For this reason, scientists must research, update, and develop them to be more perfect in helping to solve affairs, as humans can do.
Vision robot very contrary to our common sense. We can see how computer cars, which can drive automatically, often fail to detect emergencies such as pedestrians crossing the road, or where the closest object is actually when backing off.
For that, there needs to be a framework that helps this machine, as researchers from the Massachusetts Institute of Technology (MIT), United States have done. They use the system artificial intelligence (TO THE), so robot or any machine that requires vision, can analyze real-world objects from just a few images, and understand any motion when objects are used.
The research team, led by Nishad Gothoskar, a PhD candidate in electrical engineering and computer science, wrote the findings in a report paper at ArXiv, October 2021.
For optimal robot vision, they had to build a framework using probabilistic programming, an AI approach that allows the system to cross-check detected objects. The goal is that the images recorded in the camera really match, or not, with the behavior to be captured.
This programmer also allows the system to infer decisions regarding the relationship of the object being looked at, with the scene, and use reasonable reasons about inferring its position more accurately. Previously, many robots, such as AI cameras on mobile phones, had previously failed to detect how deep an object was to focus on.
Meanwhile, the installed probabilistic inference allows the system to detect when there is a possible mismatch, between caused by noise or an error in scene interpretation that needs to be corrected by further processing.
Also Read: NASA’s Rogue Robot Finds Organic Molecules on Mars
“If you don’t know about tangent relationships, then you can think of it like an object floating on a table—that would be a valid explanation. As humans, it’s clear to us that this is physically unrealistic and that the object placed on the table is a pose object. which is more likely,” explained Gothoskar in release.
“Because our reasoning system knows this type of knowledge, it can infer poses more accurately. That is a key insight from this work.”
PROMOTED CONTENT
Featured Videos
– .
