Smartphone-powered AI predicts avocado ripeness

Researchers have developed a smartphone-based artificial intelligence system that accurately predicts the maturity and internal quality of avocados.

“Avocados are among the most wasted fruits in the world due to their overripeness,” said Luyao Ma, an assistant professor at Oregon State University. “Our goal was to create a tool that helps consumers and retailers make smarter decisions about when to use or sell avocados.”

The research team, made up of scientists from Oregon State and Florida State University, trained AI models using more than 1,400 images of Hass avocados on iPhones. The system predicted firmness, a key indicator of maturity, with almost 92% accuracy, and internal quality (fresh or rotten) with more than 84% accuracy.

The results were published in the journal Current research in food science.

The researchers believe these accuracy rates can be improved as more images are added to the model. They also note that the technology has the potential to assess the ripeness and quality of other types of foods.

They hope to further develop the technology so consumers can use it at home to determine the optimal time to eat an avocado, avoiding the disappointment of cutting into one only to discover the dreaded brown spots.

The team also sees potential applications in avocado processing facilities, where the technology could be used to better sort and grade the fruit. For example, if the system detects that a batch is more ripe, it could be shipped to a nearby retailer rather than one further away. Retailers could also use this technology to determine which avocados should be sold first based on their ripeness.

These findings build on previous research using images and machine learning techniques to assess food quality. However, previous studies relied on manual feature selection. [SN1] and traditional machine learning algorithms, which limited prediction performance, said In-Hwan Lee, a doctoral student working with Ma on the project.

“To overcome these limitations, we used deep learning approaches that automatically capture a wider range of information, including shape, texture and spatial patterns, to improve the accuracy and robustness of avocado quality predictions,” Lee said.

Ma chose to focus on avocados because of their high market value and high wastage rate. She also noted a personal motivation: As an avid avocado toast consumer, she was often frustrated by not knowing when avocados were perfectly ripe and cutting into overripe avocados.

The research responds to a major global challenge: food waste. Around 30% of global food production is wasted. In response to this challenge, the U.S. Department of Agriculture and the Environmental Protection Agency have set a national goal to reduce food waste by 50% by 2030.

“Avocados are just the beginning,” Ma said. “This technology could be applied much more widely, helping consumers, retailers and distributors make smarter decisions and reduce waste.”

Zhengao Lee of Florida State University is also a co-author of the paper. Ma and Lee work in the Department of Food Science and Technology in the College of Agricultural Sciences at Oregon State. Ma is also affiliated with the Department of Biological and Ecological Engineering.