• Portada - Conference, Devoe and Blanquilla, first pear varieties for ingiGrader 2.0 training

Blog

External quality detection in pears

Conference, Devoe and Blanquilla, first pear varieties for ingiGrader 2.0 training

We have started the ingiGrader 2.0 training for pear work.

We will cover the main commercial pear varieties for the next few months. We will create a specific training for each variety since we have detected that the different colors and shades of each variety affect the estimation of the external quality of the fruit.

The system evaluates both color and near IR image captures, with the latter, defects that affect the fruit in a deeper way and are not merely aesthetic defects are detected, the near IR analysis is crucial in this type of multi-colored fruit such as: pears, apples or stone fruits.

In a few weeks the training of the system will be completed and it will be put into service in some of our customers. We already have a dataset of more than 80,000 images that will constitute the core of the system, from there we will strengthen the AI module.

Following the general philosophy of development, the external quality module will grade the fruits into 5 levels of affectation, from those corresponding to extra quality or without skin or shape defects to the fruits that, due to their high level of affectation, constitute waste or industry.