Aleksandr Shpilevoi

• Training neural networks for segmentation and object detection, tuning parameters of object trackers • Development of a Python service combining object detection, object tracking with special logic for product counting: The system makes one error for three thousand output product • Data collection, labeling validation, creation of balanced datasets, automatic data sorting, and labeling: I’ve written a Python script for auto-labeling very common defects and this saved 7000$ for the company

• I’ve integrated SORT object tracker in Hydronautics’ ROS-based “Stingray” framework • Created quick and dirty synthetic data generator utilizing blender and Python