The technologies of machine learning and computer vision (further referred to as ML and CV) are firmly embedded in technological trends. But if some five years ago only large companies could afford to use these technologies, today even a small startup can take advantage of cloud computing and deploy a project in the field of ML and CV.
The range of applications of ML and CV is continuously expanding. We can see these technologies being used to filter images, identify a person, or even help diagnose a disease. But neural networks are also widely used in different business operations.
Smart video. One of the cases of using neural networks for business is to provide help in organizing an HR video interview. If the interviewee funnel is large, an employer may ask a candidate to send a video with answers to some questions, and this might not be so easy. During the recording, you can have someone or something distracting in the background. Suffice it to recall the legendary interview on the BBC and the solution to the problem in the new version of Skype that allows blurring the background.
In one of the projects for an HR company, we at Junto have also considered other potential problems during a video interview. For example, when filming, a person does not always understand how he looks in the picture. The neural network can be trained to track the position of the head or eye movement and give recommendations to the person, making the video look more professional. Also, the software saves time to the interviewer by automatically and smoothly cutting out parts of the video where the person isn’t talking.
Sensitive data. Another application case for using CV is to work with sensitive data. In one of the projects for a bank, it was necessary to repeatedly cover up specific data in passport scans that were sent by users. By introducing an automated solution with the help of the neural network, the business was able to avoid manual retouching of scans and save on payroll costs.
Product recognition. Working with a self-service store, where a buyer takes the necessary products from the shelf, and the system automatically calculates their price, we had to implement a cost-effective solution to identify items. Initially, the business used RFID tags, that were attached to each product and a radio frequency identification scanner read the tag and understood which items the user took. Clothing stores actively work with RFID tags but using them for a grocery store proved to be too expensive, so RFID tags were replaced with Data Matrix codes. These codes are applied to the package, and a special camera reads the state of the store shelf before and after the customer approaches. As a result, the system can quickly understand what has changed on the shelf, what products the buyer took, and invoice for them.
These are just a few examples, but we hope that they gave you an understanding of how neural networks can benefit businesses, what tasks can be solved and in which areas one can see the real application of machine learning technologies and computer vision.