Transparent object bin-picking

Transparent items such as plastic bottles are notoriously difficult for traditional 3D vision systems. Glare, poor depth perception, and overlapping objects often break standard pipelines. But not ours!

 

At Automatica 2025, our system, developed in collaboration with Sedeco Imaging, combined its synthetic-data-based vision with robust perception and control strategies to showcase its capability in sorting transparent bottles. We developed a bin-picking solution capable of correctly sorting mixed soda bottles, despite variations in shape, label, and surface reflectivity.

 

What makes this work? We generate large, diverse synthetic datasets tailored to this specific application. By varying lighting conditions, bottle deformations, and labels

the robot learns in the virtual world before touching a real object. That means fewer real-world errors, faster deployment, and less reliance on expensive sensors.

 

Our goal: to make high-performance bin-picking accessible for hard-to-see objects, without redesigning your entire setup. If transparent objects are giving you trouble, let’s talk about how we can help.