Computer Vision in Python

This repository is a comprehensive, hands-on learning collection for mastering computer vision using Python. It combines well-organized Jupyter Notebooks for guided exploration with OpenCV and deep learning, alongside a practical project: a real-time finger counting application using your webcam.

🙌 Credits

This repository is based on the excellent Pierian Data computer vision curriculum, enhanced with personal projects and extensions.

📁 Repository Structure

computer-vision-in-python/
├── cv_notebook/           # Learning notebooks organized by topic
├── finger_counter_app/    # Practical application: Finger Counter
├── README.md              # You're here!

📘 Learning Notebooks

The cv_notebook folder is structured like a course and divided into the following sections:

1. NumPy & Image Basics

• Intro to arrays, image representation, and pixel manipulation

2. Image Basics with OpenCV

• Reading, displaying, and drawing on images
• Mouse interactions
• Assessments and solutions

3. Image Processing

• Thresholding, blurring, morphological operations, gradients, and histograms

4. Video Basics

• Live camera feed access
• Drawing and interacting with frames
• Video assessments

5. Object Detection

• Template matching, contour and edge detection
• Watershed segmentation
• Haar cascades for face detection

6. Object Tracking

• Optical flow, MeanShift, CamShift, and tracking APIs

7. Deep Learning for Computer Vision

• CNNs with Keras on MNIST and CIFAR-10
• Transfer learning, YOLOv3, and custom image classification

8. Capstone Project

• Finger counting using contour analysis and convex hulls

🖐️ Finger Counter App

This is a webcam-based real-time Finger Counter using OpenCV.

🔍 Features:

• ROI selection
• Background modeling with accumulated average
• Hand segmentation using image differencing
• Finger counting with contour analysis and geometric heuristics
• Live display of ROI and prediction

▶️ Run It

Ensure your webcam is connected and accessible.

python finger_counter_app/finger_counter.py

Press Esc to exit the window.

🛠️ Requirements

Install all dependencies using one of the provided environment files:

conda env create -f cvcourse_macos.yml  # or cvcourse_linux.yml / cvcourse_windows.yml
conda activate cvcourse

📂 Dataset & Assets

All necessary media and model files are included under the DATA/ and 06-YOLOv3/ directories:

• Images and videos for exercises
• Haar cascades and pretrained models
• Custom-trained Keras .h5 models for classification tasks

👁️ YOLOv3 Object Detection

A simplified YOLOv3 implementation is included under 06-YOLOv3/, adapted from the yad2k project.

• Uses darknet53 architecture
• Includes config and weights (weights need to be downloaded separately via .url file)