Azoft RnDCase Studies

Research and Development

Azoft R&D department handles the most challenging, non-standard issues that arise when implementing software solutions. Many of the problems tackled by our R&D team strongly rely on scientific research and may involve technical risks. If you have an idea or a technologically complex project but aren't sure how to make it work, Azoft R&D team can study the problem and come up with a technologically sound approach. In this section you can see some of the projects we have recently completed.

How to Detect Lung Cancer Using Artificial Intelligence

By Ivan Ozhiganov on August 17, 2017

How to Detect Lung Cancer Using Artificial Intelligence

According to the American Cancer Society, lung cancer is the leading cause of cancer related deaths. The relative survival rate for lung cancer is low – 17% for men and 24% for women. These sad statistics are a result of the large proportion of cases diagnosed at an advanced-stage of the disease. Applying AI algorithms for lung cancer detection can sufficiently improve the diagnostics quality.

Using OCR for Document Recognition and Automatic Data Entry

By Ivan Ozhiganov on July 6, 2017

OCR for Document Recognition and Automatic Data Entry

Arnold is an insurance salesman. Every single day he spends 2 hours editing reports and issuing policies. There are 18 more insurance brokers working alongside Arnold. All of them have one concern: too much manual, routine work. Since routine work is rarely exciting, it sabotages overall productivity. The time spent on typing, editing and spell checking would better be spent on client communication and skills development.

LSTM Neural Network for Accelerometer Data Processing

By Ivan Ozhiganov on April 13, 2017

LSTM Neural Network for Accelerometer Data Processing

Do you know why smartphones need embedded accelerometers? – At first, you may think it’s only for changing the screen's orientation. But the role of the accelerometer isn’t limited to just that. This small device allows you to collect data on different types of object movement. Exploiting the accelerometer data to model types of object movement requires some sophisticated algorithmic processing and machine learning techniques to cope with phone orientation changes, different type of phone accelerometer characteristics, and more. In this article, we describe how we coped with these challenges, and we present the results of training an LSTM neural network to process the accelerometer sensor data.

AI: Recommendation Systems for Retailers

By Ivan Ozhiganov on March 30, 2017

AI: Recommendation Systems for Retailers

Retailers have to compete with E-commerce shops to retain customers. Online commerce main advantages include receiving feedback quickly, opportunities to store and process data, and to answer the client requests in real time. Retail stores cannot get an immediate response from the customers, but the majority of retailers has a loyalty program and purchase information such as receipts. It's enough to create a recommendation system and provide a personalized approach to customers for increasing profits.

Information Retrieval from OCR Text

By Ozhiganov Ivan on December 14, 2016

Information retrieval from OCR text

Information retrieval is an essential stage in the analysis of camera-captured text data as there are many areas where it can be used. For example, it is highly demanded in the receipt recognition services because the quality of the recognized texts needs to be improved. To improve the effectiveness of the information retrieval from OCR texts, our R&D engineers have used LSTM recurrent neural network.

Object Detection Using Fully Convolutional Neural Networks

By Ivan Ozhiganov on November 24, 2016

Object detection Using Fully Convolutional Networks

Research on artificial neural network training is most widely used to automate multimedia data recognition. Among other things, it’s important that various types of neural networks show strong results in the tasks of speech and image recognition, financial forecasts, machine translation, and many others. We have effectively applied fully convolutional neural networks within the receipt recognition research.

Applying OCR Technology for Receipt Recognition

By Ozhiganov Ivan on April 7, 2016

Applying OCR Technology for Receipt Recognition

When it comes to the receipt recognition, it often relates to the appearance of complications with OCR methods. Extracting information from receipts implies taking a raw, badly structured text with custom characters and displaying normalized data on a device screen. Convolutional neural networks (CNN) can facilitate this process. Using CNN, we work on the comprehensive solution that can be easily adjusted to process receipts from different countries.

Convolutional Neural Networks for Object Detection

By Ivan Ozhiganov on February 25, 2016

Convolutional Neural Networks for Object Detection

Deep learning is an extremely popular field these days. Convolutional Neural Networks have successfully proven themself in computer vision. As a result, we decided to test in practice the effectiveness of convolutional neural nets for object detection in images. As the object of our research, we chose license plate and road sign pictures.

Request a Free Quote

Please enter the result and submit the form