Oakland students think up civic tech solutions to combat illegal dumping with help from locals
A group of Oakland computer science students combined forces to tackle the problem with artificial intelligence and computer vision.
OAKLAND – On a recent day on Northeastern University’s Oakland campus, an assortment of local technologists, city employees and neighborhood residents gathered under one roof to brainstorm.
The goal? Combine their diverse areas of expertise and local knowledge to generate fresh, tech-centered solutions to the age-old issue of illegal dumping in Oakland. Using the campus as a forum, the civic tech “idea-a-thon” was a chance for concerned citizens to contribute to an ongoing effort by Northeastern computer science students to tackle the problem.
The city of Oakland received over 25,000 illegal debris cleanup requests in 2025, or an average of 70 a day, according to a city audit published in April. During the last fiscal year, the city collected over 7 million pounds of trash, costing nearly $14 million. And that’s before Oakland spent an additional $2 million investigating and citing illegal dumpers, the audit revealed.
This past spring, computer science students Haoyang Li and Pranav Kishore, who spent their first year on the Oakland campus, trained two AI models to assess the scope of illegal dumping both as stationary trash piles – which can be seen in still photos of city streets – and over video as the dumping takes place. Better identification of illegal dumping sites could help the city target areas in greatest need of cleanups, Li said, leading to a more efficient system with less burden on city employees.
Joined by fellow students Alexander Tan, Naya Antaki and Henry Davis, the students trained their algorithm to spot trash piles and distinguish them from other items on the roadside, such as cars, bicycles and trash cans, Li said. Because city camera footage was not publicly available, the team simulated real city data with images created in Photoshop and from public street-view cameras in other cities, which can be found for free online.
The more difficult task was training the AI to dynamically detect, through a video feed, when someone had dumped trash through a video feed, Li said. The solution took several iterations and weeks of trial and error.
He first designed the AI to inspect a human’s arm to see if it was holding something when they entered the video frame, then check to see if that item was gone by the time they left the frame. But the task was too complex for the algorithm to handle, he said, so he instead trained it to look for items that were left in a person’s walking path by the time they left the frame.
Editor’s Picks
“One thing I learned from this process is that when you are doing computer vision, you need to think as simply as possible in how you define the action of throwing trash,” Li said. “Whether it’s trash, a bag or even a sofa, that has to be considered.”
Pranav, meanwhile, created a visual dashboard for the project with color-coded heat maps of the highest-frequency dumping sites across Oakland based on Li’s data collection.
“This gave me a chance to use some of my artificial intelligence and machine learning skills from previous experiences,” Pranav said. “It’s very rare that you get an opportunity, especially in the first year, to step away from theory and build something real.”
At the recent “idea-a-thon,” part of a weeklong series of Oakland Tech Week programming on campus, Northeastern partnered with all-volunteer civic tech nonprofit OpenOakland to continue the conversation, collecting input from the area residents most impacted by illegal dumping.
At one table, a small group of tech-minded residents hashed out possible ways to prevent dumping before it takes place. Sage Dawson, who designs video games that teach financial literacy to children, swapped ideas with Alonzo Altamirano, who works at AI software start-up Whistling Software. They were joined by Njeri Kamau-Devers, founder of Little Sheep Chinese Learning Center, an Oakland-based Afrocentric Chinese-language tutoring academy.
The three residents said they each had seen examples of illegal dumping near their homes, and came to Northeastern to offer realistic solutions. Alonzo suggested creating a decentralized archive of illegal dumping location data, which communities could use to create better-informed plans for issuing reports to the city. Kamau-Davis said forming dumping-focused group chats might help curb the issue by allowing locals to hold each other more accountable.
Over the course of the afternoon, they and other groups narrowed their ideas into actionable blueprints and presented their final proposals to Oakland civic service representatives and city employees. The concepts were catalogued for further discussion at policy panels, potential city-funded pilots or topics of future hackathons, according to Kristin Hathaway, assistant director of the Oakland Public Works’ Bureau of Environment.
Though Li and Pranav are continuing their studies in Boston, Li said he is still working toward publishing his source code for their computer-vision models. Just as Oakland attendees expressed during the recent idea-a-thon, he said it’s important that his findings be made accessible to the public so that they can be harnessed and iterated upon for the betterment of the city.
“I really love creating work that impacts a lot of people,” Pranav said. “It was a great experience working alongside people to create something that the city of Oakland could potentially use.”
