Artificial intelligence and computer vision
Augmenting computer vision by artificial intelligence has proven its ability to deal with variable products and different conditions in many industrial applications.
Flanders Make DecisionS core lab specializes on developing robust and cost-effective AI Vision systems for different industrial applications such as: objects detection, counting, pose estimation, etc.
The requirements for industrial applications are very strict: a system should work under all conditions with a very high accuracy. The application often imposes strict time requirements. Therefore, in DecisionS, we focus on developing very robust, effective solutions that are well within the computational requirements. Our algorithms are developed for the manufacturing industry, but they are often generally applicable, in other application domains such as agri-food industry.
Object detection and counting
Object detection is one of the cornerstones of AI vision. An object first needs to be detected, before it can be inspected for defects, measured, counted… In DecisionS we improved the state-of-the-art deep learning algorithms to make them more robust and predictable. Our improved version can detect tens of objects at once with high confidence. We previously applied it on industrial objects as shown in the left figure below. However, the same method with minimal adaptations could for instance be applied equally well for counting apples on a conveyor belt, as shown in the two right figures. This output could also be used to estimate the size of the apples.
Many companies want to perform visual inspection of their manufactured parts in an automated way. Our artificial intelligence algorithms are able to detect scratches, broken material and discoloration, as shown in the left figures below. With minimal adaptations, these algorithms can be applied to detect, for instance, the brown spots on apples, as shown in the right figure.
Coordinates of contact person: Abdellatif Bey Temsamani