Amazon has created three datasets to train AI systems in warehouses. The first dataset helps train AI systems to distinguish between individual products and includes over 44,000 close-up images of containers with mixed objects. The second dataset helps train AI systems to detect defects and includes over 13,300 images of defects caused by a robot picking and transferring multiple products to a destination, opening a package, or breaking a product. The third dataset helps train AI systems to direct robotic arms to separate distinct products from each other in warehouses and includes images of 235,000 order picks.
Image credit: Flickr user Dan Ruscoe