Telekinesis releases synthetic dataset for industrial bin picking on Kaggle

Telekinesis has released a new synthetic dataset specifically for industrial bin picking, now available on Kaggle. The dataset is part of a broader ecosystem that includes an "Agentic Skill library" designed to help robots learn manipulation tasks directly from simulation. By providing diverse synthetic scenes and object configurations, it enables training of computer vision models for bin picking without the cost and complexity of labeling thousands of real-world images.

The release targets the common challenge in manufacturing robotics: training AI to reliably grasp and remove parts from bins in cluttered, varying conditions. The synthetic data approach allows for scalable generation of labeled examples, speeding up deployment of vision-guided robots. The accompanying documentation and video demo show the dataset being used with the Telekinesis framework, which also integrates with ROS. For developers, this lowers the barrier to building robust industrial pick-and-place systems.