"Our goal is to offer clients smart data solutions that help them optimize business processes, adapt to a rapidly changing market, and reach a new level of business digitalization."

CEO, DUC Technologies

Optimization of solutions

Competitive advantage

Business analysis

Business values

With data and insights at your fingertips, your company can quickly make strategic decisions, adapt to the current market situation, and stay ahead of the competition.

The ability to predict in advance the software required for employees' work

Data-driven research will enable your company to recognize key trends and develop a strategy to adapt during periods of uncertainty.

1. Identified clusters of key topics related to Covid-19 in the context of business management

2. Identified key business issues caused by the pandemic

3.  Identified possible areas for further study

An analysis of Covid-19-related articles in the field of business and management in the scientific citation electronic journal Scopus

14 sources

Topic modeling scheme

A topic model in machine learning and natural language processing is a type of statistical model for identifying abstract "topics" that occur in a collection of documents.

each document consists of several topics

each topic consists of a set of keywords (terms)

All topic models are based on the same basic assumption:

Examples of research objectives based on topic modeling methods

EXTRACTION OF RELATED TOPICS FROM TEXTS

Topic modeling allows for the automatic identification of hidden themes raised in analyzed texts and the extraction of keywords that characterize each theme

RECOMMENDATIONS AND INFORMATION SEARCH

creation of a recommendation system and search engine optimization

Text filtering and classification

Automatic filtering and classification of text data based on specific topics. For example, when processing large volumes of text: news, media, etc.

CLUSTERING OF DOCUMENTS

grouping documents by their content based on identified themes to identify similar documents and search for patterns in the data

1. We selected the optimal open neural network for bag detection.

2. Prepared data for further model training

3. Retrained the model

Developing an AI-based model that can analyze video from shopping mall surveillance cameras online and identify bags and other suspicious items left behind by customers.

YOLO v8 YOLO

4. Tested on real video data

ML model diagram based on YOLO v8 YOLO

YOLO v8 YOLO (You Only Look Once) is a neural network designed for object detection in images and videos. Instead of the traditional approach of running an input image through a convolutional network multiple times to extract features and then detect objects,  YOLO performs all these steps simultaneously in a single network.

Data

Model

Prediction

1.  We used a single-class Tiny-YOLO model to detect each person in the frame.

2. Used AlphaPose to obtain the skeletal pose

3. We applied the ST-GCN model to predict actions based on every 30 frames of human tracks.

Video camera monitoring to ensure patient safety and optimize clinic operations

Tiny-YOLO

4. We created a system that not only ensures patient safety, but also helps optimize clinic operations, prevent potential incidents, and improve overall efficiency.

ML model diagram based on Tiny YOLO

Tiny YOLO (You Only Look Once) is a version of the YOLO model for object detection in images. Tiny YOLO allows for faster image processing without significantly compromising the quality of object detection.

AlphaPose is an open-source system for detecting and analyzing human poses in video with high speed and accuracy.

ST-GCN is a spatiotemporal graph convolutional network. The ST-GCN model is capable of recognizing seven human actions: standing, walking, sitting, lying down, standing up, sitting down, and falling down.