M14 / LABORATORY

OpenCV. Classical CV,
before everything went neural.

Threshold, morph, contour, transform. The toolbox that ran computer vision for thirty years still does most of the work — fast, deterministic, no weights. Eleven operations, one image, every slider live.

EXPERIMENT 07.01 · OPENCV.TOOLBOX IMAGE · circuit SIZE · 256×256 · 1ch
01 · THRESHOLD cv2.threshold
SOURCE
→
THRESHOLDED
THRESHOLD 128
MODE
02 · HISTOGRAM + EQUALIZATION
ORIGINAL
EQUALIZED
03 · MORPHOLOGY erode / dilate / open / close
BINARY IN
→
MORPHED
OPERATION
KERNEL 3×3
04 · CONTOURS cv2.findContours
EDGE THRESH 50
— contours
05 · COLOR SPACES RGB → HSV → LAB
RGB
HSV
LAB · L
GRAY
Different color spaces highlight different things. HSV separates hue from brightness. LAB is perceptually uniform.
06 · GEOMETRIC rotate · scale · translate · perspective
ROTATE 0°
SCALE 1.00
SKEW 0
07 · MASKING circular ROI
RADIUS 80
FEATHER 15
08 · EDGE DETECTION Sobel / Canny
METHOD
// 02

Source · pick the image driving all operations above

Pre-deep-learning → Every face-detector, license-plate reader, and document scanner up until ~2014 was built from operations like these. CNNs absorbed most of the workload, but classical CV is still the fastest, most predictable preprocessing layer in any vision system.