top of page

AI versus garbage

Do you have a packaging company? Do you deal with logistics? If you want to be greener, use AI (you could even save)!

Amazon has developed an AI-based system that reduces its ecological footprint. It combines product descriptions, images and structured data to decide the optimal packaging for home delivery of products. So far, it has saved roughly 2 billion boxes in packaging waste, a tremendous achievement, both economically and ecologically.

The sequence of operation:

  1. The R-CNN algorithm cuts out the product images.

  2. ResNet50, pre-trained on ImageNet, creates a separate image for the product and the manufacturer’s default packaging. For example, shipping a ball wouldn’t require a largebox.

  3. The FastText model, trained for product specifications, analyzes the text. Finding the keyword fragile, for instance, would recommend using a box.

  4. The Vanilla Neural Network generates representations of structured data—such as the number and categories of items to be shipped—to help determine if the items need packaging and, if so, how many total packages are needed.

  5. The multimodal fusion architecture combines these representations to make a packaging decision.Multimodal AI is becoming more popular, and it is encouraging to see that this approach also contributes to a more sustainable world.

5 views0 comments

Recent Posts

See All


5 (1).png

Fast. Scalable. Qualitative.

bottom of page