Hands-on labs for building computer vision applications with an AI pair programmer — from reading a single pixel array to running a deployed vision app, with Cursor, GitHub Copilot or Claude writing the OpenCV code alongside you.
📘 Course Page · 📖 Learner Guide · 🧪 All Labs · 🐛 Report Bug · 💡 Request Feature
Note
These are the official hands-on lab materials for the course:
Course Code: C592 · by Tertiary Courses / Tertiary Infotech
Course page: https://www.tertiarycourses.com.sg/ai-vibe-coding-for-computer-vision.html
1 day · 7.5 instructional hours · 2 topics · 8 hands-on labs
Lab 1 — Set Up Your AI Vibe Coding Environment · Isolated Python environment, OpenCV, and your first script written from a prompt rather than from memory.
Lab 2 — Load, Display and Transform Images · Resize, crop, rotate, flip and colour-convert — and deliberately trigger the BGR/RGB bug that turns skin blue.
Lab 3 — Filtering, Thresholding and Edge Detection · A four-panel comparison view with live Canny trackbars, because parameter tuning is where AI assistants are least reliable.
Lab 4 — Contours: Find, Count and Measure Objects · The full pipeline to a labelled count and a CSV export. The first run gives the wrong answer on purpose.
Lab 5 — Detect Shapes and Faces · Shape classification from contour geometry, then Haar-cascade face and eye detection with no training data and no GPU.
Lab 6 — Object Detection with a Pre-Trained Model · YOLOv8 detection with confidence and class filtering, and a per-class summary count.
Lab 7 — Working with Video Streams · Webcam and file capture loops, frame-differencing motion detection, annotated MP4 output, and detection on every Nth frame.
Lab 8 — Build, Test and Run Your Computer Vision App · A Streamlit app with real error handling and a pytest suite over the processing functions.
This repository contains the complete lab materials for AI Vibe Coding for Computer Vision (C592) by Tertiary Courses / Tertiary Infotech. Each lab is a self-contained activity that gives you the exact prompt to paste into your AI coding assistant, the code it should produce, and a "Test it" step that tells you what a correct result actually looks like.
The premise of the course is that an assistant which produces runnable code is easy; one that produces correct code needs you to know what the right answer looks like. So every lab has a verifiable outcome — a count that must match, a shape that must be named correctly, a box that must land on the object — rather than "it ran without an error".
| # | Lab | Concepts |
|---|---|---|
| 1 | Set Up Your AI Vibe Coding Environment | venv, OpenCV install, assistant setup, prompt structure, imread returning None |
| 2 | Load, Display and Transform Images | NumPy image arrays, resize with aspect ratio, ROI crop, rotate/flip, BGR vs RGB |
| 3 | Filtering, Thresholding and Edge Detection | Gaussian blur, fixed vs Otsu threshold, Canny, trackbars, convolution kernels |
| 4 | Contours: Find, Count and Measure Objects | findContours, RETR_EXTERNAL, area filtering, moments/centroids, CSV export |
| 5 | Detect Shapes and Faces | approxPolyDP, circularity, Haar cascades, scaleFactor/minNeighbors, ROI search |
| 6 | Object Detection with a Pre-Trained Model | YOLOv8, boxes/labels/confidence, class filtering, COCO class limits |
| 7 | Working with Video Streams | VideoCapture loops, release discipline, frame differencing, VideoWriter, every-Nth-frame inference |
| 8 | Build, Test and Run Your Vision App | Streamlit UI, pure functions for testability, failure cases, pytest with synthetic images |
📖 Full walkthrough: see the Learner Guide for detailed step-by-step instructions for every lab. Slides, the Learner Guide and the Lesson Plan are in
courseware/.
| Category | Technology |
|---|---|
| Vision Library | OpenCV (opencv-python) — image and video processing |
| Arrays | NumPy — every image is an ndarray |
| AI Pair Programmer | Cursor, GitHub Copilot or Claude |
| Object Detection | Ultralytics YOLOv8 (yolov8n, CPU inference) |
| Classical Detection | Haar cascades bundled with OpenCV (cv2.data.haarcascades) |
| App Framework | Streamlit — upload, process and download in the browser |
| Testing | pytest with synthetic NumPy fixtures |
| Plotting | Matplotlib — used to demonstrate the BGR/RGB trap |
| Editor | VS Code or Cursor |
| Courseware | Slides (python-pptx), Learner Guide + Lesson Plan (python-docx) |
TOPIC 1 — Getting Started and Image Processing with AI Vibe Coding
Lab 1 prompt ─▶ assistant ─▶ hello_cv.py ─▶ imread ─▶ imshow (the vibe coding loop)
Lab 2 images/ ─▶ resize · crop · rotate · flip ─▶ outputs/
└─▶ matplotlib preview ─▶ BGR vs RGB
Lab 3 image ─▶ greyscale ─▶ GaussianBlur ─▶ threshold / Otsu
└─▶ Canny (trackbars) ─▶ edges_best.png
Lab 4 image ─▶ greyscale ─▶ blur ─▶ Otsu(INV) ─▶ findContours
─▶ area filter ─▶ moments ─▶ counted.jpg + objects.csv
TOPIC 2 — Detection and Building Vision Apps with AI
Lab 5 contours ─▶ approxPolyDP + circularity ─▶ shape names
image ─▶ grey ─▶ Haar frontalface ─▶ faces ─▶ ROI ─▶ Haar eye
Lab 6 image ─▶ YOLOv8n ─▶ boxes + class + confidence
─▶ --conf / --classes filter ─▶ per-class summary
Lab 7 VideoCapture(0 | file) ─▶ frame loop ─▶ raw | edges | motion
─▶ VideoWriter ─▶ output.mp4
└─▶ YOLO every Nth frame ─▶ cached boxes
Lab 8 upload ─▶ app.py (Streamlit UI) ─▶ vision_ops.py (pure functions) ─▶ result + download
└─▶ test_vision_ops.py (pytest)
C592-AI-Vibe-Coding-for-Computer-Vision/
├── README.md
├── LG-AI Vibe Coding for Computer Vision.md # Full step-by-step learner guide (start here)
├── screenshot.png # Real pipeline output (Labs 3, 4, 5)
│
├── labs/ # All hands-on labs (one folder each)
│ ├── README.md # Lab index
│ ├── resources/ # Shared sample images
│ │ ├── shapes.png # 5 shapes — known-correct answer
│ │ └── objects.jpg # 12 objects + grit — 41 raw contours
│ ├── lab01-set-up-your-ai-vibe-coding-environment/
│ ├── lab02-load-display-and-transform-images/
│ ├── lab03-filtering-thresholding-and-edge-detection/
│ ├── lab04-contours-find-count-and-measure-objects/
│ ├── lab05-detect-shapes-and-faces/
│ ├── lab06-object-detection-with-a-pre-trained-model/
│ ├── lab07-working-with-video-streams/
│ └── lab08-build-test-and-run-your-computer-vision-app/
│
└── courseware/ # Slides, Lesson Plan + Learner Guide
├── AI Vibe Coding for Computer Vision-v1.0.pptx # 93-slide deck (+ PDF)
├── LG-AI Vibe Coding for Computer Vision.docx # detailed step-by-step (+ PDF)
└── LP-AI Vibe Coding for Computer Vision.docx # 1-day lesson plan (+ PDF)
- A Windows or Mac laptop with at least 8 GB RAM and permission to install software
- Python 3.10+ — installed directly or via Anaconda
- VS Code or Cursor
- An AI coding assistant: GitHub Copilot in VS Code, Cursor's built-in assistant, or Claude in a browser tab
- A webcam for Lab 7 (any short MP4 works instead)
- (Optional) a Google Colab account as a fallback if a local install fails
git clone https://github.com/tertiarycourses/C592-AI-Vibe-Coding-for-Computer-Vision.git
cd C592-AI-Vibe-Coding-for-Computer-Visionmkdir cv-vibe
cd cv-vibe
python -m venv .venv
.venv\Scripts\activate # macOS/Linux: source .venv/bin/activate
pip install opencv-python numpy matplotlibThis prints an OpenCV 4.x version and opens a small black window. Click it and press any key to close.
python -c "import cv2, numpy as np; print('OpenCV', cv2.__version__); cv2.imshow('ok', np.zeros((120,240,3), np.uint8)); cv2.waitKey(0); cv2.destroyAllWindows()"Copy the sample images across and open the first lab:
mkdir images
cp ../labs/resources/*.jpg ../labs/resources/*.png images/Then follow labs/lab01-set-up-your-ai-vibe-coding-environment/README.md.
⚠️ Never close an OpenCV window with the X button — click the window and press any key, or the Python process hangs. Every lab's script ends withcv2.waitKey(0)andcv2.destroyAllWindows()for this reason.
💡 Read the generated code before you run it. The point of this course is that you stay in control of the logic — not that the assistant is always right. Labs 4, 5 and 8 deliberately produce a wrong-but-runnable result so you learn to spot one.
For complete, step-by-step setup, see the Learner Guide.
Contributions, fixes, and improvements are welcome:
- Fork the repository
- Create a feature branch:
git checkout -b feature/my-improvement - Commit your changes:
git commit -m "Add my improvement" - Push the branch:
git push origin feature/my-improvement - Open a Pull Request
Found a bug or have an idea? Open an issue.
This material is provided for educational use as part of the course C592 — AI Vibe Coding for Computer Vision. © Tertiary Infotech Pte. Ltd. All rights reserved.
Tertiary Infotech Pte. Ltd. — Tertiary Courses Course: AI Vibe Coding for Computer Vision (C592)
- OpenCV — the computer vision library the whole course is built on
- Ultralytics — YOLOv8 pre-trained detection models
- Streamlit — turning a Python script into a usable app
- Cursor, GitHub Copilot and Claude — the AI pair programmers used throughout
- Course trainers and learners of C592
⭐ If this helped you learn computer vision, star the repo!
Powered by Tertiary Infotech Academy Pte Ltd