Coursework for the AGH Advanced Vision Systems (AVS) class. Labs cover classical CV fundamentals (image ops, background subtraction, optical flow, filtering, tracking); two projects apply those tools end-to-end.
| Folder | Topic |
|---|---|
lab_1 |
OpenCV basics — image I/O, drawing, color spaces |
lab_2 |
Background subtraction via frame differencing (pedestrian sequence) |
lab_3 |
Background subtraction with mean / median frame buffers (highway, office) |
lab_6 |
Image filtering with SciPy / NumPy |
lab_7 |
Optical flow — block-matching baseline |
lab_8 |
Multi-object tracking with SiamFC appearance model |
project |
Toothbrush bristle defect detection. Classify SEM / high-res bristle-tip images as defective or clean. |
project_2 |
EVS-MOT pedestrian tracking. Multi-object tracking on the EVS-MOT challenge dataset using BoT-SORT + a ConvNeXt-Small ReID encoder. Two pipeline variants — yolo-version/ (own YOLOv8 detector + ensemble options) and dettxt-version/ (uses the challenge-supplied det.txt). |
Each folder has its own requirements.txt / pyvenv.cfg.
Larger model weights are tracked via Git LFS.