Amazon wants to get rid of the barcode and will use artificial intelligence to achieve it

Amazon wants to get rid of barcodes and… uses artificial intelligence to do so. The company developed an AI model that uses a camera to identify products so that it will no longer be necessary to scan a barcode. According to the e-commerce giant, this system will streamline the processing of shipments in distribution centers, which will result in faster delivery times.

Whereas barcodes are fundamental to shipping processes Amazon wants to eliminate reliance on them. There are multiple reasons to do this, but the main one is automation. The website digitizing products requires an employee, as robots are not versatile enough. to handle a product and manufacture it.

When an item arrives at an Amazon logistics center, employees use barcodes to verify its identity at several different locations along its journey to the delivery vehicle. Each time, the item must be picked up and the barcode located and scanned. Sometimes the barcode is damaged or even missing.

To modernize its warehouses, Amazon has shared details of the process known as . multimodal identification (MMID)). The first step in eliminating the reliance on barcodes was to take pictures of the products as they move along the treadmill. The AI ​​model takes values ​​such as dimensions, visual features, packaging text, or weight.

Amazon has created an AI model to identify its products and eliminate reliance on barcodes

Amazon AI
Amazon AI
Amazon AI

Using photographic cameras and depth cameras. a sort of fingerprint of each object is created. The researchers then translated the data from each image into vectors and built a machine learning model to extract it and associate it with the product to be compared.

According to Amazon, the match rate of the algorithm stayed between 75% and 80% when first used. Through the constant capture of new images to train the AI ​​model, the MMID achieved 99% accuracy. According to the engineers, the high match rate is also due to the fact that the company’s inventory systems know where each item is, so the algorithm does not need to match one product to another. entire company catalog.

The system is designed to be non-intrusive and is able to detect an error quickly.. The individual trays are part of the initial phase of the process, so if there is an anomaly it is resolved without having to wait for the end. The model uses a trust rating system which determines if there is a mismatch or if the algorithm is not fully secure.

The MMID depends on several factors to ensure correct operation. First, the cameras take photos which feed the database and train the algorithm. The lighting and speed of the treadmill are also important, and Amazon makes sure they are controlled. The website the only inconvenience occurs when an employee handles the object in the tray. It makes detection more difficult depending on how it is held.

Amazon Engineers are now working to integrate it into robotic arms so that the presence of a human is no longer necessary in the future. The company hopes to eliminate reliance on manual item identification, a process that is cumbersome and inefficient.

We would like to say thanks to the writer of this post for this remarkable web content

Amazon wants to get rid of the barcode and will use artificial intelligence to achieve it

You can find our social media profiles here as well as other pages on related topics here.