What we do
We conduct research in various areas of computer vision:
- Object Class detection using trained models – Cascade Classifier and Support Vector Machines (SVM);
- Single object detection, using perspective transformations and combinations of local feature detectors and classifiers;
- Object and scenes detection using Convolutional Neural Networks (CNN);
- Detection of various complex object classes with big variations of appearance and structure, consisting of limited number of parts;
- Human identification using Microsoft Kinect and Eigen Faces;
- Video Stream segmentation on scenes (video file or camera capture);
- Image grouping and classification using visual similarity approach;
- Video clips grouping and classification on visual similarity.
Sirma Computer Vision Lab is a research unit of Sirma Group Holding, focused on development of technologies for computer interpretation of 2D and 3D image data sets and video streams. The laboratory conducts research in the areas of object detection and recognition, human detection and motion analysis, OCR, scene recognition, material recognition, visual measurement etc.
The main objective of the lab is to create products and technologies which would have real world applications in diverse industries such as manufacturing, media and entertainment, healthcare, security and robotics. As a part of an IT business group, we strive to develop top-notch technologies in different domains, which works successfully and deliver great value to their customers.
We are currently developing our OCR Toolkit, optimized for text recognition of images, captured from smartphone digital cameras. Our software achieves state-of-the-art results for low-quality images that are very challenging for other OCR SDK libraries.
Our advanced pre-processing technology allows correction of typical degradations, specific to camera-captured images, such as uneven lighting, insufficient illumination, 3D perspective distortions, blur, noise, curved text lines.
There are lot of applications that will benefit from the improved OCR recognition quality, as document scanning, text extraction from receipts, business cards, etc.
We have started work on a project related to human movement analysis, which we believe, will become the next revolution in computer vision, and will introduce a myriad of opportunities for advancement in the areas of computer games, security and healthcare.