AI REGIO - Supervised real-time 2D-based Object Detection System
A flexible tool able to identify and localize in Real-time the best object to pick in scene with a multitude of overlapped identical objects.
Due to the immense variability in terms of dimension, geometry and materials that typically can be found in an industrial scenario, a modern approach to machine vision must be applied especially for the majority of objects, mostly comprised of small parts, typically characterized by symmetric geometry. This 2D object recognition and localization pipeline approach is able to deliver faster inference speed compared to other detectors with a good accuracy level. The inference speed of the algorithm is an important factor from an industrial perspective because it has a direct impact on the cycle time. This algorithm could find different applications in industry, a typical case is its use in a robot perception system for bin-picking operations. The asset has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 952003 (AI REGIO).
Among the most important object detectors belonging to one-stage methods, this SSD detector achieves a fine balance between speed and accuracy. The algorithm achieves good results in terms of speed and accuracy (reaching a mAP of mAP@0.5 IOU 0.85, mAP@0.75 IOU =0.57) for the detection of small components in a scene.
A custom variant of the algorithm performs well also for the detection of an object among a multitude of identical objects also overlapped.