Automating the fruit quality control process

The company estimated the yield of each harvest through a manual process carried out by a group of expert inspectors, whose evaluations were based on a sampling rate of approximately 2 out of every 1,000 fruits. This limited the inspection capacity and made it difficult to obtain representative information for estimating production yield.

Additionally, the company needed to strengthen its quality control processes to meet the stringent standards required by international markets while obtaining more accurate information to support harvest planning and packing operations.

We implemented a computer vision MVP by installing a fixed camera on a packing line and developing YOLOv3-based detection models to automatically analyze fruit characteristics during the inspection process.

The solution integrated image processing into the existing operational workflow, establishing a digital mechanism for capturing and analyzing fruit quality data. It also laid the foundation for automating harvest yield calculations and generating digital reports based on the collected information.