AI adventures in arts and letters – TechCrunch

More news Possibly be with anyone there. But you can bear the most interesting developments with this column, which gathers AI and machine learning advances from around the world and explains why they may be important for technology, startups, or civilization.

To start on an eighth note: machine learning methods for researchers learning art are always interesting – though not always practical. A team at the University of Washington wanted to see if a computer vision system could learn to tell what was being played on the piano with an overhead view of the keys and the player’s hands.

Oudo, A system coached by Eli Schleizerman, Kun Su, and Schulong Liu, watches a video of the piano playing and first pulls out a simple sequence of key presses like a piano-roll. Then it combines the expression as to the length and strength of the press, and polishes it for input into the MIDI synthesizer for final output. The results are a bit loose but definitely recognizable.

A video showing the piano player's hands on the keys turns into MIDI scenes.

Image Credit: Schleiserman, et. al

“To make music that sounds like it could be played in a musical performance was previously thought impossible,” Schleizman said. “An algorithm needs to detect cues, or ‘features’, to generate music in a video frame, and it needs to ‘visualize’ the sound occurring in the middle of the video frame.” This requires a system that is both precise and imaginative. The fact that we acquired music that sounded great was a surprise. “

This is another fascinating research from the field of art and letters. Computational disclosure Very delicate to handle ancient letters. The MIT team was looking at “closed” letters of the 17th century that have been bent and sealed so intensely that removing and leveling the letter could permanently damage them. His approach was to set up a new, advanced algorithm to X-ray the papers and decrypt the resulting imagery.

A diagram showing the X-ray views of a letter and how it is analyzed to virtually reveal it.

A diagram showing the X-ray views of a letter and how it is analyzed to reveal it exactly. Image Credit: MIT

MIT’s Eric DeMinen said, “The algorithm does an impressive job of separating layers of paper and despite the occasional small gaps between them.” “We were not sure it would be possible.” This task can be applied to many types of documents that are difficult to open for simple X-ray techniques. It is a bit of a stretch to classify it as “machine learning”, but it was very interesting not to include it. Read the full paper Nature communication.

The diagram showing the review of electric car charge points has been analyzed and transformed into useful data.

Image Credit: Asensio, et. al

You arrive at a charge point for your electric car and put it out of service. You can also leave a bad review online. In fact, thousands of such reviews exist and constitute a potentially very useful map for municipalities to expand electric vehicle infrastructure.

Omar Asensio of Georgia Tech Training a natural language processing model Such reviews and it soon became an expert in parsing them by thousands of people and squeezing out insights where outages were common, comparable costs and other factors.