Time-Lapse Videos Through Guided AR

An app developed by Cornell researchers uses augmented reality to help users repeatedly capture images from the same location with a phone or tablet to make time-lapse videos – without leaving a camera on site. Time-lapse photography, which involves combining photos taken over long periods of time, provides a powerful way to visualize phenomena such as changing seasons or the movement of the sun. Traditionally, photographers would leave a camera on a tripod for the duration of the event, but researchers in the Cornell Ann S. Bowers College of Computing and Information Science, have developed a more convenient method. Their iOS app, ReCapture, is now freely available in the Apple app store.

The researchers believe this is the first application designed for creating time-lapse videos from handheld devices. The app has three capture modes that cover a range of scenarios. One works best for landscapes, one helps capture close-up scenes and a third collects a range of images that can be used to reconstruct the scene in 3D offline. Each capture mode uses different information about the scene. The simplest mode uses an overlay of previous shots to help the user line up new photos. For close-up scenes, which tend to be more difficult to capture, the application tries to figure out where the camera is in 3D space and uses arrows to tell the user how to move and tilt their phone toward the correct location.

More information:

https://news.cornell.edu/stories/2022/11/app-creates-time-lapse-videos-smartphone

Low-Cost Obstacle Robot

Scientists at Carnegie Mellon University and the University of California, Berkeley, have enabled a low-cost and relatively small legged robot to adapt to obstacles. The robot uses its vision and an onboard computer to quickly adjust to new situations and master difficult terrain. The researchers trained it using 4,000 robot clones as they walked and climbed in a simulator, giving the machine six years of experience in one day.

The simulator also retained motor skills acquired in training in a neural network that the team copied to the actual robot. The team put the robot through its paces, testing it on uneven stairs and hillsides at public parks, challenging it to walk across steppingstones and over slippery surfaces, and asking it to climb stairs that, for its height, would be akin to a human leaping over a hurdle. The robot adapts quickly and masters challenging terrain by relying on its vision and a small onboard computer.

More information:

https://www.cmu.edu/news/stories/archives/2022/november/visual-locomotion.html

AI Trains Robot Dogs

Carnegie Mellon University researchers used an artificial intelligence (AI) to train a robot dog to perform cleaning tasks for less than a tenth of the cost of Boston Dynamics' robot canines. The researchers taught the AI to coordinate the robot's movements with an arm affixed to its back while an operator guided its activity. 

 

They applied reinforcement learning to train the AI via computer models and in a physical machine. The researchers trained the AI to direct the robot's legs separately from the arm, before training it on leg/arm coordination. The team also had a teacher AI train a student AI to mimic their bodily motions.

More information:

https://www.newscientist.com/article/2344914-having-ais-train-robot-dogs-to-balance-makes-them-a-lot-cheaper/