OpenCV Python Workshop

Hands-on OpenCV workshop using Python. Covers image processing, feature detection, face recognition, video analysis, object detection, segmentation, and edge deployment.

Workshop by Dr. Farshid Pirahansiah — www.tiziran.com | YouTube

Setup

Quick Install (2025-2026)

# Create virtual environment
python -m venv cv_workshop
source cv_workshop/bin/activate  # Linux/Mac
cv_workshop\Scripts\activate     # Windows

# Install dependencies
pip install -r requirements.txt

Legacy Install (from notebooks)

pip install numpy pandas matplotlib
pip install opencv-python opencv-contrib-python
pip install Pillow bokeh seaborn

Contents

File	Description
opencv_functions.py	Reusable utility functions (cartoon, face detection, display)
opencv_python_visualcode.py	VS Code integration examples
list_files_directories.py	File/directory listing utilities
mat2numpy.py	OpenCV Mat to NumPy array conversion
read_all_image_in_folder.py	Batch image loading
progress_bar.py	Progress bar utility
yolo_detector.py	YOLOv11 real-time object detection
segmentation.py	SAM-2 style image/video segmentation
video_analyzer.py	Motion detection, tracking, optical flow
edge_deploy.py	ONNX export, INT8 quantization, benchmarking
augmentation.py	Image augmentation pipeline (Albumentations + OpenCV)
features.py	SIFT/ORB/AKAZE, feature matching, panorama stitching
haarcascades/	Haar cascade classifiers for face/eye detection
lbpcascades/	LBP cascade classifiers (faster, lighter)
tests/	Pytest test suite

New Modules (v1.1)

YOLO Detector (yolo_detector.py)

from yolo_detector import YOLODetector

det = YOLODetector(confidence=0.25)
result = det.detect("photo.jpg")
for d in result.detections:
    print(f"{d.class_name}: {d.confidence:.2f} at {d.bbox}")

# Video/webcam
stats = det.detect_webcam(camera_index=0)
stats = det.detect_video("traffic.mp4", output_path="annotated.mp4")

# Export to ONNX
det.export_onnx("yolo11n.onnx")

Segmentation (segmentation.py)

from segmentation import ImageSegmenter

seg = ImageSegmenter("sam2_n")
result = seg.segment("photo.jpg")
overlay = ImageSegmenter.visualize(cv2.imread("photo.jpg"), result)
cv2.imwrite("segmented.jpg", overlay)

# Interactive: click foreground/background points
result = seg.segment_interactive("photo.jpg")

# Video segmentation
stats = seg.segment_video("video.mp4", output_path="seg_out.mp4")

Video Analyzer (video_analyzer.py)

from video_analyzer import VideoAnalyzer

analyzer = VideoAnalyzer("video.mp4")
result = analyzer.detect_motion(method="mog2", min_area=500)
print(f"Motion frames: {result.motion_frames}/{result.total_frames}")

# Object tracking
result = analyzer.track_objects(tracker_type="csrt")

# Optical flow
result = analyzer.compute_optical_flow(method="farneback")

Edge Deployment (edge_deploy.py)

from edge_deploy import EdgeDeployer

deployer = EdgeDeployer("yolo11n.pt")

# Export
onnx_path = deployer.convert_to_onnx("yolo11n.onnx")

# Quantize
result = deployer.quantize_int8("yolo11n.onnx")
print(f"Compression: {result.compression_ratio:.2f}x")

# Benchmark
bench = deployer.benchmark("yolo11n.onnx", num_runs=200)
print(f"Latency: {bench.avg_latency_ms:.1f}ms, FPS: {bench.throughput_fps:.1f}")

# Jetson deployment
deployer.deploy_jetson("yolo11n.onnx", "jetson_deploy/")

Augmentation Pipeline (augmentation.py)

from augmentation import AugmentationPipeline

pipeline = AugmentationPipeline()
pipeline.create_transforms(rotate_limit=45, color_jitter=True)

augmented = pipeline.augment(image, num_variations=5)

# Batch augment directory
pipeline.augment_directory("images/", "augmented/", num_variations=3)

# Visualize pipeline
vis = pipeline.visualize_pipeline(image, num_samples=8, output_path="vis.jpg")

Feature Detection (features.py)

from features import FeatureExtractor

extractor = FeatureExtractor()

# Extract features
kp, desc = extractor.extract_sift(image)

# Match two images
result = match_result = extractor.match_images(img1, img2, method="sift")
vis = extractor.draw_matches(img1, img2, result)

# Panorama stitching
panorama = extractor.panorama_stitch([img1, img2, img3])

Topics Covered

Fundamentals

• Image I/O (read, write, display)
• Color space conversion (BGR, HSV, GRAY)
• Geometric transformations (resize, rotate, warp)
• Drawing and annotations

Image Processing

• Filtering (Gaussian, median, bilateral)
• Thresholding (adaptive, Otsu)
• Morphological operations
• Edge detection (Canny, Sobel, Laplacian)
• Histogram equalization and analysis

Feature Detection & Matching

• ORB, SIFT, AKAZE feature extraction
• BFMatcher and FLANN matching
• Lowe's ratio test
• Homography estimation (RANSAC)
• Panorama stitching

Object Detection

• Haar/LBP cascade face/eye detection
• HOG pedestrian detection
• Contour-based object detection
• YOLOv11 real-time detection with batch support

Segmentation

• SAM-2 interactive and auto-segmentation
• Mask visualization and polygon extraction
• Video frame-by-frame segmentation

Video Processing

• Webcam capture and processing
• Background subtraction (MOG2, KNN)
• Optical flow (Farneback, Lucas-Kanade)
• Object tracking (CSRT, KCF, MOSSE)
• FPS and performance statistics

Deep Learning & Edge Deployment

• ONNX Runtime inference in Python
• INT8 dynamic quantization
• Benchmark comparison (latency, memory, throughput)
• TensorRT export hints
• Jetson deployment package generation

Data Augmentation

• Rotation, flip, crop, color jitter
• Albumentations integration with OpenCV fallback
• Batch augmentation from directories
• Dataset export with CSV labels

Modern Python CV Stack (2025-2026)

Library	Use Case
opencv-python	Core image processing and CV
ultralytics	YOLOv11 detection/segmentation
onnxruntime	Cross-platform DNN inference
albumentations	Advanced image augmentation
supervision	Video annotation and tracking
mediapipe	Face/hand/pose landmarks
torchvision	PyTorch vision utilities

CLI Usage

Each module can be run standalone:

# YOLO detection
python yolo_detector.py photo.jpg --conf 0.3 --output result.jpg
python yolo_detector.py video.mp4 --output annotated.mp4
python yolo_detector.py 0  # webcam

# Segmentation
python segmentation.py photo.jpg --interactive
python segmentation.py video.mp4 --output seg_out.mp4

# Video analysis
python video_analyzer.py video.mp4 --mode motion --motion-method knn
python video_analyzer.py video.mp4 --mode track --tracker csrt
python video_analyzer.py video.mp4 --mode flow --flow-method farneback

# Edge deployment
python edge_deploy.py yolo11n.pt --convert --quantize yolo11n.onnx
python edge_deploy.py --benchmark yolo11n.onnx --num-runs 200

# Augmentation
python augmentation.py images/ --num 5
python augmentation.py photo.jpg --visualize

# Feature matching
python features.py img1.jpg img2.jpg --method sift --output matches.jpg
python features.py img1.jpg img2.jpg img3.jpg --stitch --output panorama.jpg

Resources

• OpenCV Python Tutorials
• Ultralytics YOLO
• ONNX Runtime Python
• Albumentations
• YouTube Channel

12-Month Roadmap (2025-2026)

Month	Milestone	Status
Jul 2025	Python 3.10+ migration, type hints, pathlib	Done
Aug 2025	Pytest test suite >80% coverage, CI/CD	Done
Sep 2025	Docker multi-stage builds, GPU support	Done
Oct 2025	YOLOv11 integration, ONNX Runtime examples	Done
Nov 2025	SAM 2 segmentation, real-time inference	Done
Dec 2025	Edge deployment guide (Jetson, Raspberry Pi)	Done
Jan 2026	INT8 quantization, benchmark comparison	Done
Feb 2026	Feature matching, panorama stitching	Done
Mar 2026	Video analysis (motion, tracking, optical flow)	Done
Apr 2026	Augmentation pipeline (Albumentations + OpenCV)	Done
May 2026	Multi-camera system examples, 3D reconstruction	Pending
Jun 2026	v1.0 release, comprehensive documentation	Pending

Version History

Version	Date	Changes
1.1.0	2026	Add YOLO detector, segmentation, video analyzer, edge deploy, augmentation, features
1.0.0	2025	Full modernization: Python 3.10+, type hints, pathlib, pytest, Docker
0.x	2019-2024	Initial workshop materials

License

See repository for license details.