Data annotation can Help Streamline Online Shopping

Computer vision based AI models are transforming the online shopping experience. Companies, like Amazon and eBay, that facilitate consumer to consumer selling amass enormous online inventories, featuring billions of individual listings. A consequence of the ever increasing size of online shopping catalogues is increasing complexity for consumers.

Finding the right item, posting new listings, and avoiding fraudulently priced products can be challenging for users of many online shopping inventories. Advancements in machine learning are beginning to be deployed to streamline these processes, making listing and finding items secure and convenient. This development is supported by the work of data annotation.

By adding labels to images of thousands of objects annotators create robust training datasets that form the basis of many AI models.

This blog will identify the challenges facing customers online, and show how AI systems are being developed to overcome these issues, with the help of data annotation services.

Existing challenges

Ever expanding online inventories present a number of problems for consumers:

• Finding the right item: It can be hard to track down specific items on websites. If a customer sees a specific item that they like in the real world or elsewhere online it can mean searching through thousands of similar (or not so similar) objects before a close match can be found. Potential sales can be lost due to the time and effort required to find exact matches for target shopping items.
• Creating new listings: For some individual sellers creating listings represents a significant investment of time and effort. Sellers that sell many types of one object class may have to create hundreds of individual listings, entering information each time. This delay can result in loss of earnings for sellers and frustration for potential buyers.  
• Identifying problem items: At certain times particular item types can be subject to fraudulent selling practices and price inflation. During the COVID-19 pandemic, high demand items like face masks and hand sanitizers have been routinely marked up in price, leading to desperate customers being illegally overcharged. Other items may be in violation of local laws. However, the volume of items being listed can make problematic ones hard to pinpoint and remove without automation.

Computer vision solutions

The recurring issues highlighted above continually affect the online shopping experience. As a result many companies are turning to computer vision AI models to find solutions:

• Simplified searching: Traditional searches use word terms that may not capture the specifics of the item being searched for. Computer vision allows customers to find more relevant alternatives by analysing images at a pixel level. AI models can be trained to identify images taken by consumers and correlate them with objects in online inventories. This process allows shoppers to quickly search using images from other sources.
• Streamlined listings: Item listing can also be greatly accelerated by machine learning powered AI applications. Object recognition technology allows online shopping apps to recognise items and automatically create listings. This allows sellers to create a high volume of listings whilst maintaining a high quality of attached information.
• Automated item checking: AI can help retailers to target fraudulent and exploitative selling practices. Computer vision models are able to recognise target objects, like hand sanitizer, en masse. They can then identify cases in which these items are being intentionally over priced. This means that problem sellers can quickly be removed from any platform.

Managing data with the annotation services

These AI use cases promise to improve online shopping for both sellers and buyers. However, accessing effective image annotation services is still an important factor for developers hoping to improve the functionality of their computer vision systems. Keymakr offers competitive pricing for AI innovators, in combination with a managed, in-house annotation team that guarantees precise annotation and management flexibility.