Using Data Annotation to Make AI in Retail Possible

Category: AI insights

Published date: 31.07.2020

Read time: 6 min

AI is expanding into virtually all industries and retail is no exception. In fact, experts are forecasting that the spending on AI in retail will be $12 billion by 2023. The reasons for this are quite simple: AI offers a boost in efficiency for retailers while lowering costs. The customers are offered a greater level of convenience since AI streamlines many processes as we shall see later allowing people to find the products they are looking for faster and expedite the checkout process as well. However, in order to make all of the cutting-edge products possible, a lot of data annotation work is required in order to train the machine learning algorithms to function properly.

With all of this in mind, let’s take a look at some of the practical applications of AI in retail and the data annotation techniques behind them.



Self-checkouts are starting to replace human workers and we are seeing a lot of investments coming in to make them more user friendly and increase their functionality as well. In fact, the self-checkout market is expected to topple $5 billion by 2024. These include traditional systems where people simply scan their items and put in their bags to more extreme use cases like what we are seeing with stores like Amazon Go. Companies like Bingo Box have followed suit opening up more than 500 stores across China. Both Amazon Go and Bingo Box are fully automated stores without a single human employee, the only difference is the products they offer.

The reason that someone can just walk into an Amazon Go store, take something off the shelf, and simply walk out is because of facial recognition technology. The AI system matches the customer’s picture with what they have on file and simply bills the credit card they have on file. Human data annotators place landmarks around thousands and thousands of facial images for the computers to understand all of the aspects of the human face. For example, they need to know what eyebrows, noses, lips, and all of the other facial features look like, so it can recognize them in virtually every image. With high-tech stores like Amazon Go, a lot of facial annotations would be required given that the enterprise itself depends on this technology working properly.

Automated Warehouses

AI is starting to replace humans in warehouse management as well paving the way for a greater level of efficiency. In the old days, companies needed to hire somebody to physically walk around the warehouse with a pen and clipboard and make sure all of the carets and supplies were in the right place along with the necessary quantity. AI along with computer vision has changed all that. For the purposes of this discussion, let’s just take a look at one aspect of warehouse management: travel time i.e. walking around the warehouse to find the needed item. This process alone eats up around 50% of the total picking time in warehouses and fulfillment centers around the world. Imagine how much time and money such a routine and basic process is costing companies.

AI-powered robots are helping companies solve these issues by automating these redundant, yet necessary processes. A great example of this is a company called Ocado, a British online grocery retailer, that uses robots to fulfill all of the orders. Inside their warehouses, robots literally roam around, pick up and pack the groceries. These robots rely on LiDAR technology to recognize their surroundings which also requires data annotation. LiDAR produces a 3D point cloud, which is a depiction of how the robots see the world. Humans need to annotate it by labeling all of the objects. Some advanced forms of annotation include color-coding the distances of the items in relation to the robot with blue being the lowest and red the highest. This technology allows robots to move around all by themselves without any human assistance.

Also, the robots must be able to identify all of the objects in the warehouse in order to pack them. This means that humans would need to annotate all of the images of the items by tagging, 2D/3D bounding boxes, or semantic segmentation to make sure the right items are delivered to the customers. The more images are annotated, the higher the accuracy will be. While it is possible to use historical data sets, you need to keep in mind that the packaging and other features of the product change over time, so it would be better to use fresh data for your machine learning project.

Mindy Support Can Facilitate the Development of Your AI Project

Mindy Support has extensive experience actualizing all kinds of data annotation products. Our size and location allow us to source and recruit candidates to assemble even the most sizeable teams within a short time period. If you are operating under tight deadlines, we can help you keep the development of your project on schedule. Our rigorous QA processes ensure that all of the work will be done right the first time. Get the best quality-price offer in the market. Check the quality yourself with our free trial. Contact us today to set a call and discuss your project.

Posted by Il’ya Dudkin


    Stay connected with our latest updates by subscribing to our newsletter.

      ✔︎ Well done! You're on the list now