ɪnˈtɛɡr(ə)l dx, PhD
  • Home
  • Projects
  • Research

Aerial Object Tracker

2026-01-22 0 min read Page Views

A drone view dataset is used to fine-tune the base Faster RCNN model to learn aerial object detection.

https://i.imgur.com/g2DzkRx.gif

The model is further enhanced with object tracking. Moving objects are retained using epipolar constraints. A trick is to assume that all objects not satisfying epipolar constraints are moving objects.

https://i.imgur.com/oLoEcng.gif

Last updated on 2026-01-22
Previous
Teaching LLM to play chess ... badly
Next
Dots: surround and capture, in Rust
Back to Home
© 2026 ɪnˈtɛɡr(ə)l dx, PhD | Powered by hugo-theme-void | Total visits: