VisionSystems - AI-based software development studio - Portfolio

VisionSystems
Data Science studio

AI-based software development. Machine learning & Computer Vision

Portfolio

Projects we delivered for our customers

AI-based restaurant key human operarations analytics

Restaurant operations videoanalytics

Restaurant business is highly dependent on quality and efficiency of cooking, serving and cleaning operations. Our customer, a chain of fast restaurants wanted to put core operation under automated unbiased control. For this we have developed software that tracs key operations. It identifies employees and detects operation they are performing at each moment, tracks number of people in the que for making order and how soon tables and are cleaned after guests finished their meal. A central web-portal provided visual analitics to business management helping them to make better business-decisions.

CV and sample source code evaluation automation

AI-based intelligent IT-candidate rating system

ML-based system that evaluates job applications and quality of computer code samples submitted for software developer positions. This technology multiplies the efficiency of hiring process. It allows HR managers and recruiters to hire faster and more accurately processing more applications and consuming much less time of expensive software experts for professional computer code evaluation.

Recommender System for e-commerce: how to help your customer find what he really needs?

Shopping recommender system

VisionSystems have designed and developed a complex recommender system for the North American online shopping platforms. It captures data from the customer's personal details, preferences, past and current shopping sessions, purchases of the other customers similar to him and transform it into dynamic offers. It then scans through the entire product catalogue and lines up the best products for the individual customer, which allows to increase sales and customer satisfaction.

Human emotion recognition and analisys basing on Computer Vision techniques

Intelligent service quality control

There is always a flow of positive and negative customers at the service front-line. Our client - a chain of pizza restaurants was not an exception. The task of the project was to track the customers' emotional state and to measure service satisfaction rate within the work of specific attendants and restauraunts.

We have used our own, custom-made framework - Happybox, a boxed solution, that uses computer vision and machine learning to automatically analyze service quality and customer's emotions based on the the video data.

Computer Vision system detecting ATM user age, sex, emotions and gestures

Smart ATM concept

The request of our client (regional buank in Europe) was to develop the concept of of Smart ATM (HappyBot). Smart ATM was supposed to detect visitors sex and age, identify emotions, gestures and invite visitor in interactive games.

The prototype of the ATM was demonstrated on the Bank's Booth at The Forum of Innovative Financial Technologies - Finopolis 2019 and obtained popularity among participants. The ATM could also identify all photos in the event photobank containing visitors face and emailed him selected photos.

Computer vision based detection of currency exchange operations

Cash fraud detection

Many of the banks face the same old problem - cash fraud. Some cashiers do not register the currency exchange operations in the bank's cash desk. Instead they simply use their own money to exchange the clients' amounts and put the profit from exchange rate difference in their pocket.

As a result the bank receives less profit and bears reputational risks.

We have developed cash detecting system (ExchangeGuard), which is able to detect the types of currency in the tray by analyzing video stream and to compare these events to the records stored in bank's automated transaction system to reveal fraudulent activities. In case systems containes no record of the visually identified currency exchange transaction, security/quality control department is notified and video fragment containing operation is attached to such notification of immediate verification.

Comuter Vision and NLP making recorded online meetings searchable and analyzable

Face recogniton system for videoconferences

Developed for customer special needs, our face recognition tool runs on the customer servers' side within security perimeter. The product functionality includes:

flexible adjustment of the employees list, based on their names and positions, both within the entire company and for each department separately
automated and manual recognition of specific virtual room participants
addition of unrecognized employees tho the employee lists "on-the-fly"
a frame face recognition and their signature.

We have closely cooperated with the in-house development team and ensured successful and easy integration of the new face recognition functionality into an operating high-load system product.

Computer vision detects object states and presence to notify attendant

Livestock feed availability control system

Feeding cows in dairy farms is the most critical process that affects the amount of milk. We have developed the system detecting the forage quantity on the feeding table of dairy farms for one of our customers.

The constant availability of fresh forage in sufficient volume is extremely important, since it allows to obtain the highest milk yield, but special forage for these needs is too expensive to simply be constantly replaced. Hence, the quantity of the feed must be strictly calculated. Moreover, each cow has it's own feeding behaviour and it is impossible to predict distribution of forage during the day with sufficient accuracy.

Using the existing customer video surveillance infrastructure, our system enables to detect the forage availability and quantity and sends signals to the farmers when a new delivery is required.