04 / RESEARCH / COMPUTER VISION
Kinemation.
Making motion make sense. Turning human movement into coherent, frame-to-frame skeletal animation.
Pose Estimation Research Intern · ACM AIEP
Python · MediaPipe · OpenCV · NumPy
THE STARTING POINT
A frame is only part of the story.
Pose estimates can look accurate one frame at a time and still produce jittery animation. Kinemation investigates how temporal context can turn those individual estimates into coherent movement.
THE APPROACH
From question to implementation.
- 01
Worked on a real-time human pose estimation pipeline using MediaPipe and OpenCV during the ACM Industrial Experience Program.
- 02
Researched temporal cohesion techniques to improve stability between consecutive frames, rather than treating each pose as an isolated prediction.
- 03
Explored 3D pose estimation to extend the pipeline beyond 2D skeletal tracking.
- 04
Contributed within the Kinemation research team. The shared repository documents the wider pipeline, including keypoint adaptation and temporal smoothing approaches.
WHERE IT LED
What came out of it.
Research experience connecting computer vision papers to an implementation, with a focus on temporal consistency. Internship period: January–April 2026. The team repository tracks ongoing research and implementation.
Explore the code and documentation