Amazon releases the largest dataset to train \”pick and Place\” robots
Amazon released the largest dataset ever of images taken in a product-sorting industrial setting. Amazon’s ARMBench dataset contains more than 190,000. Whereas the previous largest dataset of industrial images contained on the order 100 objects, this dataset from Amazon, called ARMBench features over 190,000. It could also be used to \”train\” robots to better generalize their abilities to different products and contexts.
In a paper that we will be presenting later this spring, at the International Conference on Robotics and Automation.
In the scenario where ARMBench was used, a robotic hand must pick up a single object from a bin of objects and place it on a tray. This task is made more challenging by the variety of objects, their configurations, and their interactions within the context of a robotic system.
Source:
https://www.amazon.science/blog/amazon-releases-largest-dataset-for-training-pick-and-place-robots