We've worked with big enterprises and medium-sized companies from the USA, Europe and Asia. That’s why we can deliver high quality and reliable products matching your business needs.
We offer end-to-end Machine Learning, Data Science, AI & software development. We will take care of your entire project from the inception phase to scalable deployment to production.
Risky and irresponsible driving is a worldwide problem - over 1 million people die in road crashes every year. Financial costs of these crashes is estimated at over 2% of GDP.
Our goal is to build systems that detect dangerous maneuvers on the road, help both people to improve their driving skills and businesses to reduce fleet management costs.
We are creating a deep learning-based platform that analyses images captured by a dashboard camera and detects dangerous events, such as tailgating or lane hopping. We have collected and annotated a data set, created and trained a deep learning model and deployed it on the mobile phone. The inference is done on the edge, while analytical and scoring algorithms are run on the backend.
We are constantly adding new types of recognised dangerous events to further improve the precision and scope of our analysis.
When it comes to security user experience usually suffers. Remembering passwords or carrying keycards around the office is cumbersome. This often leads to people by-passing the security measures. We wanted a secure and easy way to authenticate our workers to our back-office software system.
We have employed state-of-the-art deep learning strategies to create a module that recognizes company employees based on an image of a face. We have implemented a triplet loss algorithm on top of a deep convolutional neural network. We have first pretrained the model on a large set of publicly available face images and then applied transfer learning technique to fine-tune the model to our specific use-case.
Technical or scientific documents usually have a lot of figures, tables or diagrams besides text. In one of our systems it was crucial to extract these kind of objects from a PDF document as a separate image and take note of its location in the original document.
We have used deep learning techniques to identify objects of interest in the documents. We have implemented the YOLO detection algorithm and evaluated multiple architectures of the underlying convolutional neural network. We have pretrained the model on a large set of labeled images and then applied transfer learning to our specific task. To increase the data set for our specific task, we have applied data augmentation.