Hybrid Agrobots

Manual egg inspection is slow, subjective and hard to staff consistently. Hybrid Agrobots needed software that could inspect produce at line speed — catching dirty, cracked and blood-spotted eggs, and flagging under- and over-weight ones against each customer’s own criteria — accurately enough to trust, fast enough to keep pace with a conveyor running 40,000 eggs per hour, and adaptable to other round produce in future.

Indivar, as Hybrid Agrobots’ software partner, built BoldVision: a C++ application that runs the full inspection pipeline in real time. A 12-megapixel camera captures multiple images per second of each egg as it passes; a trained machine-learning model — built on current state-of-the-art vision models and running on an NVIDIA Jetson Nano edge module — classifies surface defects such as dirt, cracks and blood spots, while a load cell measures weight live and checks it against the customer’s under- and over-weight thresholds. BoldVision then drives the segregation hardware: out-of-spec eggs are diverted into dedicated lanes — dirty, cracked-or-blood, underweight and overweight — while perfect eggs continue through to packing. The same vision-and-weight engine is designed to grade round fruit and vegetables — apples, oranges, lemons and more — by retraining the model.