Dmitrii L.

Linkedin: [login to view URL] Github: [login to view URL] Have experience in: - Implementation Lidar clusterization algorithms: DBSCAN, Ransac, ABD (Adaptive Breakpoint Detector), DBD (Dual Breakpoint Detector), algorithms from PCL Library (Euchlidean etc.); - Implementation of track and train detection CV algorithms ( using a grid ) and a multi-class neural network; - Implementation of multi-class algorithms for detecting objects on a image using YOLO v3 open dataset; - Rewriting Python Tensorflow code to C++; - Writing an API on Qt C++ ( [login to view URL] ) for connecting with Livox Lidars such as Mid-40/100, Horizon, Tele-15. - Improving positioning when moving around the map. Completion of an electronic map for the movement of autonomous locomotives. - Development of a system for simulating weather conditions and anomalies for subsequent recognition by neural networks; - Development of image classification system; - Improvement of image quality in frames from a video pair; - Analysis of object detection methods on images in existing projects; - Adding new functionality in the form of applying new object recognition methods using OpenCV and Cuda libraries for C++ language; - Writing utilities to expand the dataset of the neural network (simulating weather conditions, highlighting the bounding boxes on images); - Processing 8-bit image masks from the neural network to determine the rail tracks; - Writing a transcoder to convert data from old formats to new; - Data fusing from various sources (neural networks, classical methods of computer vision, tracks from the navigation map); - Implementation of algorithms for recognizing weather conditions and anomalies; - Implementation of a module for diagnosing a disparity map from a stereo pair; - Smart Map Matching algorithm implementation using a probabilistic approach and navigation electronic database (sqlite + C++) for reduction of latitude error; - Particle Swarm Filter implementation for reduction of longitude error; - Involved in the implementation of a Kalman Filter; - Porting Kalman filter from ROS to the train requirements; - RailSLAM algorithm implementation using navigation data and binary masks from neural network; - Drawing up requirements for navigation, localization and an electronic database for locomotive positioning; - Writing technical documentation for the customer; - Participation in writing a common company code style; - Analysis of object detection methods on images in existing projects; - Adding new functionality in the form of applying new object recognition methods using OpenCV and Cuda libraries for C++ language;