This repositoty contain an end to end application that receive image of detection stick, and return Positve/Negative status for the detection.
A - YOLOv8 based Object Detection Model: Locate the stick and the exact window were markers are expected. Then croping the test window ans aligning it, so that it is directed horizontally.
B - Markers Detection - Running Classical CV methods, as pattern match, to enhance pixels were markers can be found.
C - Classification - Light CNN classification model. Given the map of the potential markers, and their intensity, classifying the test to be positive or negative.
| Positive | Negative | |
|---|---|---|
| source |  |
 |
| alligned window |  |
 |
| markers |  |
 |
| model response | Positive, confidence = 1.0 | Negative, confidence = 0.92 |
Please follow instructions in INSTALL.md