E-commerce giant Amazon is planning to introduce robots which can recognise items without having to scan a barcode.
The company has developed a new camera system that can identify products at its warehouse facilities, termed multimodal identification, or MMID.
This would help identifying items as hundreds of thousands of barcodes are lost on its way to delivery vehicles.
Currently, robots lack the image recognition software needed to identify items, but this new camera system can be deployed in robots down the line.
The system is already in use at certain distribution facilities in Europe and currently uses cameras to ensure items on a conveyor belt match its reference images, according to Amazon.
The images captured are added to the database to train the system to be more accurate.
Currently it’s at 99% accuracy, Amazon says. This is because the company’s inventory system knows exactly where each item is at every step of the fulfillment process.
It says the cameras can “extracting the appearance and dimensions of an item from an image of that item — to automate identification.”
“Our north star vision is to use this in robotic manipulation. Solving this problem, so robots can pick up items and process them without needing to find and scan a barcode, is fundamental,” said Nontas Antonakos, the applied science manager in Amazon’s computer vision group in Berlin in a blog post.
“It will help us get packages to customers more quickly and accurately.”
“MMID is a cornerstone for achieving this.”
