91麻豆天美 researchers honored internationally for AI-powered wood identification research
Contact: Chloe Madison
STARKVILLE, Miss.鈥擜n international wood society is recognizing 91麻豆天美 researchers with a prestigious award for groundbreaking work harnessing artificial intelligence to identify wood species.
The team, led by Frank Owens, associate professor in 91麻豆天美鈥檚 Department of Sustainable Bioproducts, is the recipient of the George Marra Award from the Society of Wood Science and Technology which was presented at the organization鈥檚 recent conference in Portoroz, Slovenia. Owens and collaborators, including Assistant Professor Adriana Costa, Department Head Rubin Shmulsky, and Team Leader of the USDA Forest Products Laboratory鈥檚 Center for Wood Anatomy Research Alex Wiedenhoeft, were recognized for their 2023 article 鈥,鈥 published in Wood and Fiber Science.
鈥淓very year, the Society of Wood Science and Technology gives the George Marra award to authors that demonstrate excellence in research and writing,鈥 Owens said. 鈥淚 know and respect many of SWST鈥檚 members, so it鈥檚 gratifying that our work was recognized by a peer group as worthy of an award.鈥
Owens, Costa and Shmulsky, who are also scientists in the university鈥檚 Forest and Wildlife Research Center, are pioneers in leveraging this technology to advance the field of wood identification.
鈥淎round 2010, research on using computer-vision, a type of artificial intelligence, for wood identification started. As researchers rushed to gather data and publish, we noticed the quality of wood specimen surface preparation varied from image to image and publication to publication,鈥 Owens said. 鈥淪ometimes the knifing or sanding quality was so bad you couldn鈥檛 see the features needed to identify the wood specimen. Our team had always considered high-quality surface preparation to be essential for training and testing accurate models, figuring that if we humans couldn鈥檛 see the wood features adequately, maybe the computer couldn鈥檛 either. No one had really tested that assumption. We decided it was important to investigate further.鈥 聽
The team captured thousands of cross-sectional images of tropical woods under magnification to train the model. They tested the model on images of wood specimens prepared across progressively coarser sanding grits鈥攆rom flawless to extremely scratchy. They were the first group to identify the predictive accuracy of the model drops as the coarseness of the sanding increases.
Costa, who focused on preparing samples and capturing the thousands of images needed to train the model, said it鈥檚 an honor to receive the award during her first year as an assistant professor.
鈥淚 always tell my students to be patient with their craft,鈥 she said. 鈥淓very project will have its challenges, and sometimes when you鈥檙e solving a problem, it will lead to a new project. Keep going, work ethically and don鈥檛 take short cuts.鈥
Other collaborators on this project included University of Wisconsin Model Developer Prabu Ravindran; Assistant Professor Brunela Pollastrelli Rodrigues of Clemson University, who sanded and imaged the testing specimens; and Manuel Chavesta and Rolando Montenegro of Universidad Nacional Agraria La Molina, who provided expertise on the tropical wood species used in the project.
To learn more about the Department of Sustainable Bioproducts in 91麻豆天美鈥檚 College of Forest Resources, visit . To learn more about the Forest and Wildlife Research Center, visit .
91麻豆天美 University is taking care of what matters. Learn more at聽.