AI and Artificial Vision: Automatic Container Identification




Efficient merchandise storage system is one of the key points for the Supply Chain and Reverse Logistics Management. Most of the companies face the multiple challenges caused by the lack of warehouse and/or open land storage area, absence of effective picking and storage procedures, and insufficient or unstructured information about the storage elements. To top up this list of operational dares, the companies should deal with limitations constituted by the nature and characteristics of the most commonly used storage instruments – containers.

In 1994 Denso Wave, a subsidiary of Toyota, invented QR Codes (history) to improve the manufacturing process of vehicles and its parts. QR Codes information storage represented a 300-fold increase with respect to Barcodes that allowed to upgrade the tracking procedure of the containers with the product inside. But reverse logistics poses its own unique challenges, that could be solved with the help of Computer Vision and AI solutions. 

Resource optimization for logistics company

For a large-scale transport and logistics company, to solve the problem of storage of its empty containers is a top priority and the biggest issue, as it manages hundreds of containers of different shapes and sizes within a short period of time. What makes it even more complicated is that once the merchandise is delivered to the customer, containers lose part of the information encrypted in the QR Code used to track the goods that were inside. Yet, the empty containers must be temporally stored within the limited space of the warehouse until the reverse logistic process is finished. Management of these containers must be carried out based on the containers shape and the partial information that’s left.

Normally, to group the same types of containers in one place, operators must go through the company’s warehouses to place the former where they belong, wasting time and money on doing so. It was a strategic need for the client to improve such a resource consuming method by using the tools operational optimization and principles of connected logistics.  

The objectives set for Aquiles team were clear and answered one question: Is there an AI system that can recognize containers based on its shape and that will allow to maintain real time container inventories in the reverse logistics process?

In other words, the application should be able to:

  • Automate container class identification
  • Reduce unnecessary forklift travel
  • Improve control of the number of containers in the area


The bespoke solution for this challenge consisted in developing a computing system based on the Artificial Vision. Aquiles team suggested to design an AI tool that would automatically recognize the container that is to be loaded on a forklift and immediately indicate the exact location in the warehouse where the container should be moved and placed to the operator.

Challenges of inventory management in reverse logistics

A neat solution of the AI recognition system, managed with a help of computer vision, required solving a list of implications:

Dealing with large volume and variety

When offering logistic and storage service to a major player of the vehicle manufacturing industry, the company manages in the range of hundreds of container types, meant to carry different parts and equipment that will later participate in the construction of a vehicles in the assembly line. These containers are made of different materials, metal, plastic or expanded polystyrene (EPS) to name a few. Therefore, a broad categorization should be made in order to draw a more subtle difference among them and get a more straightforward optimization system.

Optimal storage solutions

Products, reusable containers in this case, tend to remain longer in reverse channels than in forward channels, resulting in higher costs in inventory, transportation and warehousing. Given a large quantity of the containers, their variety and frequent turnover, the storage locations are temporally and usually happened in ill structured spaces. This fact sets an extra AI optimization challenge that, once solved, can bring a significant improvement into the operational processes.

Timely and accurate inventory control

Finally, real time inventories of the containers in the reverse logistic process must be maintained in order to develop a near-optimal logistics system. Neglecting it would end up into a part-time effort approach and a decrease of the efficiency of the Supply Chain that will worsen with time.

How to avoid common pitfalls of the reverse logistics and how to safeguard the large quantity of the reusable material in a correct and orderly manner?

Automatic container identification

At Aquiles Solutions we love challenges, and this was a tough one. There is no public Image Data Base to train an AI system. Building one was a first step in the software development and it set quite a challenging operational quest.

Phase 1: Taking information and developing the tool

An embedded computing system was developed and installed onto forklifts to harvest containers video recording as drivers would run their daily operations. Usability was paramount to minimize interference with the drivers’ duties. And, for that reason, alongside with a simple board computer and camera, a tablet for a straightforward introduction of the necessary information was installed on the operating forklift.

Phase 2: Designing prototype system

Developing agile video tagging tools ensured the creation of a broad data set to train our designed AI systems. A prototype system was designed and introduced for real situations testing. By using advance AI image recognition, Aquiles team was able to combine partial information left after poorly classified containers from the different forklifts, retrained and improve the model, thus, boost AI container recognition and meet the client’s needs.

Phase 3: Managing reverse logistics with AI image processing

Once containers were identified, the AI system needed to assign the optimal temporary storage for the container until the reverse logistic process is completed. The tool was created to communicate with the company’s warehouse management system providing the type of container that was detected, to obtain the location of the container and show it to the operator. AI based optimization techniques proved to be the perfect answer for a real-time reverse logistics inventory considering retrieval time, storage time or rotation.

AI image recognition and advance optimization techniques are two building blocks that constitute the main elements of a real-time reverse logistics inventory.


Using AI for reverse logistics offers a multitude of solutions to get the world closer to the ideal of a circular economy. AI-driven image processing, Data Science, and other digital techniques simplify reverse logistics by optimizing the operational processes.

In this case, managing reverse logistics with the help of Computer Vision and AI, the logistics and transportation company achieved the following benefits:

  • Keeping track of all containers in the inverse logistics process.
  • Reducing operations cost and time by using AI optimization techniques.
  • Gaining flexibility to include new containers on-the-go and have it monitored by the AI system right away.


Reverse logistics sets its own challenges since most of the tracking technology is meant to assist the monitoring of the goods involved in the manufacturing process and forward logistics. Using AI for image recognition and optimization of the storage and supply chain procedures can help fulfill those operational dares that all major companies face. Recognizing the container during the reverse logistics process in the logistics sector gives a bright example of how AI solutions empower the on-going objectives of optimizing the operations, reducing time and minimizing costs.


More cases

Process Partner: Anomaly Detection in manufacturing
Smart Roundabout based on Deep Reinforcement Learning
Purchase Order Reader: Automated Data Extraction