CREATE Lab’s Yen-Chia Hsu trains volunteers to label smoke at GASP HQ.
Congratulations are in order for Carnegie Mellon University’s CREATE Lab and project leader Yen-Chia Hsu on the successful launch of its smoke reading tool – one that can label industrial emissions.
The project involved months of work – six setting up the infrastructure and 12 more working with volunteers to refine the system and label videos depicting industrial emissions in an effort to train an artificial intelligence system to detect smoke automatically.
GASP this past fall was proud to partner with CREATE Lab on this project, hosting training sessions and helping to enlist volunteers.
Last week, CREATE Lab released the data and code on GitHub, a platform typically used for hosting open-source projects. A technical report was also published.
“We have developed a computational model to automatically recognize smoke emissions, using the data that you and others contributed,” Hsu told GASP. “For the next step, I will use the model to visualize air pollution events in Pittsburgh.”
GASP Executive Director Rachel Filippini lauded the work, saying that it will be beneficial for local air quality.
“Seeing is believing,” she said. “Being able to visualize local air pollution events will not only help people better understand emissions sources, but it will also add another layer of accountability to industrial polluters and regulators alike because the proof will be right there for all to see.”