Machine learning
Advantages
- Experience in developing custom algorithms.
- Development of software based both on our own algorithms and on customer's algorithms.
- Close cooperation with the professional mathematical community.
Core technologies
- Real-time Signal Processing Algorithms
- Standard and specialized image and video stream preprocessing algorithms
- Compression algorithms for various types of data streams, including images
- Use of color spaces, interpolation and color image processing
- Edge Detection and Filtering in an Image Segmentation Problem
- Description of different types of boundaries and algorithms for combining them
- Segmentation of various objects, ranging from simple (linear area, spatial area, small size) to detection of human faces
- A set of algorithms for solving recognition problems (binarization, segmentation, border thinning, morphological analysis, expansion, erosion, filters)
- Recognition: construction of space attributes, classification of objects, self-learning of the recognition system
- Automatic testing and tuning of classification algorithms using image databases
- Detection and tracking of objects in the video stream in real time
- Development of a 3D model of objects from a series of frames
- Construction and processing of depth maps using binocular systems
- Evaluation of the metric characteristics of moving images of objects and their dynamics by a series of frames
- Development of algorithms for data visualization, including 3D graphics
- Implementation based on various platforms: x64, GPU (OpenCL, Cuda), FPGA
- Development of all stages of the algorithm: formulas, Matlab, x86, built-in solution
- Real-time Signal Processing Algorithms
- Standard and specialized image and video stream preprocessing algorithms
- Compression algorithms for various types of data streams, including images
- Use of color spaces, interpolation and color image processing
- Edge Detection and Filtering in an Image Segmentation Problem
- Description of different types of boundaries and algorithms for combining them
- Segmentation of various objects, ranging from simple (linear area, spatial area, small size) to detection of human faces
- A set of algorithms for solving recognition problems (binarization, segmentation, border thinning, morphological analysis, expansion, erosion, filters)
- Recognition: construction of space attributes, classification of objects, self-learning of the recognition system
- Automatic testing and tuning of classification algorithms using image databases
- Detection and tracking of objects in the video stream in real time
- Development of a 3D model of objects from a series of frames
- Construction and processing of depth maps using binocular systems
- Evaluation of the metric characteristics of moving images of objects and their dynamics by a series of frames
- Development of algorithms for data visualization, including 3D graphics
- Implementation based on various platforms: x64, GPU (OpenCL, Cuda), FPGA
- Development of all stages of the algorithm: formulas, Matlab, x86, built-in solution
This system is capable of recording high-resolution video from two interconnected cameras, correcting the resulting image, and even predicting possible emergencies. All this would not have been possible without the development of the latest algorithms and the use of mathematically complex software.
Development of an algorithm for Audio and Video recognition software for the hotel and restaurant market segment This project uses an individual approach to the client using patented algorithms During the implementation of the project, an in-depth study of the subject area of the customer was carried out.
This project is the development of software for accurate calculations. The main task was to develop a modern graphical interface and optimize the used computer resources. This method is used to optimize various calculations and allows you to:
- carry out preliminary calculation;
- cache intermediate results;
- perform parallel computing.