Food, drink and retail businesses throughout the trendy Baltimore neighborhood of Hampden took major hits when the coronavirus pandemic forced closures. Starting Monday, many of those unemployed and underemployed workers will get a chance to work on a project that helps astronomers better search images of outer space.
The Hubble Image Similarity Project is an initiative of the Baltimore-based Space Telescope Science Institute, which operates the science behind the famed Hubble Space Telescope.
Joshua Peek, an associate astronomer at the institute, said the project grew out of a desire to better search its immense database of images. So far, he said, scientists typically search metadata associated with the images to find similar photos taken by the Hubble Telescope, a model of which sits in the institute.
Peek, who has a background in both astronomy and image processing, wanted to use neural networks — algorithms that seek out existing patterns — to improve upon the existing methods.
“The problem is, the neural networks that exist right now are really good at looking at a picture of a chair and seeing that it is a chair and not a cat, or a cat and not a dog, or whatever," Peek said. “But we don’t have that information for space.”
Peek’s research requires other eyes to make sure that images are being categorized correctly. Recognizing this need, he reached out to the nearby Hampden neighborhood to seek out volunteers who could tell whether certain images of galaxies and stars looked like other images of space.
Conversations with area restaurateurs led him to Lou Catelli (real name: Will Bauer), who is informally called the “mayor of Hampden.” Catelli quickly began recruiting people through community organizations and businesses, as well as social media outlets like the Hampden Neighbors Facebook page.
“My job was super easy: just connecting everybody [involved] quite quickly, because they’re going to ramp up quick,” Catelli said.
It works like this: Workers log onto the website that Peek and his team set up, click on similar images when prompted, and receive about $20 an hour for their work. Payment is sent through Amazon Mechanical Turk, a crowdsourcing marketplace, and the team hopes to pay around $2,000 to each worker over the next month or two.
Peek said he’ll start contacting participants Monday. One of those waiting for possible confirmation, Peter Treadway, learned about the opportunity from Catelli. Treadway works as a line cook at Nepenthe Brewing Company, close to the Avenue, and lost shifts due to the closures.