Daniil Osokin
Chief Programmer at Octavian Spb
Belgrade, Central Serbia
Summary
🤩
RockstarDaniil Osokin is a Chief Programmer based in Belgrade with 11 years of software engineering experience, specializing in machine learning and real-time computer vision. He is the primary author and maintainer of lightweight human pose estimation projects in PyTorch, focusing on CPU-efficient, real-time multi-person 2D and 3D pose estimation and pragmatic deployment optimizations (OpenVINO, TensorRT). Daniil’s contributions span algorithm refinement, robustness improvements like pose smoothing and ID propagation, and deployment tooling such as conversion scripts and pure-Python demos that make research-ready models easier to run in production. He combines an engineer’s attention to performance with practical usability improvements, and his work is notable for enabling fast inference on resource-constrained environments.
11 years of coding experience
English
Github Skills (11)
tensorrt10
computer-vision10
pytorch10
deeplearning-ai10
deep-learning10
realtime10
python10
human-pose-estimation10
machine-learning8
openvino7
mlops7
Programming languages (6)
Github contributions (5)
Daniil-Osokin/lightweight-human-pose-estimation-3d-demo.pytorch
Oct 2019 - May 2021
Real-time 3D multi-person pose estimation demo in PyTorch. OpenVINO backend can be used for fast inference on CPU.
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 13 commits, 12 PRs in 1 year 6 months
Contributions summary:Daniil primarily contributed to the project by enhancing its machine learning capabilities. They added a pure Python demo for pose extraction, which could indicate development of or experimentation with pose estimation methods. Moreover, they integrated TensorRT support for optimized inference, including conversion scripts and adjustments to the inference engine, showcasing an interest in model deployment and performance optimization. The user also fixed a pose-skipping issue.
pytorchcpuperson-pose-estimationpersondeep-learning
Daniil-Osokin/lightweight-human-pose-estimation.pytorch
Mar 2019 - Jul 2021
Fast and accurate human pose estimation in PyTorch. Contains implementation of "Real-time 2D Multi-Person Pose Estimation on CPU: Lightweight OpenPose" paper.
Role in this project:ML Engineer
ML EngineerContributions:1 review, 35 commits, 30 PRs in 2 years 5 months
Contributions summary:Daniil contributed to the project by implementing and refining human pose estimation functionality. They addressed issues such as GPU availability and added features like results preview and pose ID propagation, improving the system's usability. Further enhancements included code cleaning and performance optimizations, particularly in the demo and pose-related modules. They also introduced pose smoothing techniques to improve the robustness of the pose tracking.
multi-personreal-timecpuhuman-poseperson
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.