He has near 8 years of experience in Machine Learning, Ph.D. in Mathematical Modelling and Numerical Methods (01.05.02). His main field of interest is Neural Networks for Computer Vision tasks.
Object detection, segmentation and pose estimation for mobile devices
Data Science and Machine Learning
Object detection, segmentation, and pose estimation are the most popular tools for ML mobile applications. However, there are a lot of approaches, papers, repositories for each of these tasks. Firstly, it is hard to choose an optimal one. Secondly, even harder to combine several such repositories into a single solution. Thirdly, there are a lot of problems to deploy these approaches to the embedded devices.
We consider simple effective methods to solve all these tasks, describe in detail how to better optimize loss function during training, help to choose operations and layers to deploy on an embedded device and show performance on Mobile.