Computer vision
Three cameras. One shared plane.
I worked on a mask/no-mask detector and projected detections from three camera views onto a common top-view plane.
My role
I built the course project with Suleman Mahmood.
Where it stands
LUMS computer-vision course project. Recorded detection scores used training images; they do not establish held-out accuracy. Tracking and duplicate removal were not implemented.
What changed.
A YOLOv5s detector connected to a manually calibrated, three-camera homography prototype.
Connect detection to geometry
We labeled video frames, trained a two-class detector and projected detection centroids into a shared floor plane. Four manually selected correspondences defined each camera's homography.
The limits are part of the demonstration
The cameras required manual calibration and the views used simple blending. The dataset was small and the recorded evaluation reused training images, so I present the pipeline and its limitations rather than a deployment-ready accuracy claim.
The stack
- Python
- YOLOv5s
- PyTorch
- OpenCV
- Homography
- ffmpeg