Amazon has revealed details of Project P.I. (Private Investigator), utilizing computer vision techniques to scan products for damage before packaging them for customer delivery. This helps reduce product returns and customer complaints.
Products in Amazon’s warehouse are sent through an image tunnel for scanning purposes. Project P.I. then analyzes various defects such as book covers that are wrinkled and alerts human staff to decide on the best course of action for problematic items, such as selling them as second-hand goods or donating them.
Amazon states that Project P.I. solves issues related to customer satisfaction and sustainability by eliminating the need to return products, reducing packaging waste and carbon emissions from transportation.
Originally, Amazon used OCR technology to read text labels on products and compare them with database information. However, the limitations of this technique only allowed for text recognition. With the advancement of computer vision technology, the company can now analyze data in more detail, creating learning models on how “correct” products should appear. If there are photos of products that look different, the model is able to differentiate between them.
Next, Amazon plans to introduce Multi-Modal LLM (MLLM) to enhance Project P.I. in cases where products pass AI’s scrutiny but still end up with customers. If customers complain and report issues, reviews from customers will be analyzed and compared with the images captured by P.I. to determine the cause of the discrepancy, such as incorrectly attached product labels by staff. This ensures that the distribution center can address and correct label issues to prevent future errors.
TLDR: Amazon’s Project P.I. uses computer vision to identify product defects before packaging, reducing returns, and complaints, with plans to implement Multi-Modal LLM for further enhancement in handling customer issues.
Leave a Comment