Modern Computer Vision Repository

This repository contains a comprehensive collection of Jupyter notebooks covering a wide range of computer vision techniques and deep learning applications for image and video processing. It serves as both a learning resource and a reference implementation for various computer vision tasks.

Repository Structure

The repository is organized into two main directories:

1. Deep Learning CV

Contains notebooks focused on deep learning approaches to computer vision, including:

• Convolutional Neural Networks (CNNs)

  • CNN Visualizations (Filters, Activations, GradCAM)
  • Transfer Learning with pre-trained models
  • Performance analysis and model fine-tuning

• Image Classification

  • Cats vs Dogs classification with PyTorch
  • Fashion-MNIST implementations with regularization techniques

• Generative Models

  • GANs (Generative Adversarial Networks)
  • CycleGAN for image translation (Horses to Zebras)
  • ArcaneGAN for artistic style transfer
  • StyleGAN for anime generation

• Image Segmentation

  • DeepLabV3
  • Mask R-CNN
  • U-Net and SegNet architectures

• Object Detection

  • YOLOv3, YOLOv4, and YOLOv5
  • Faster R-CNN implementations
  • SSD (Single Shot Detector) with MobileNetV2
  • Custom detectors for specific applications (Chess pieces, maritime objects, potholes)

• Web Applications

  • Flask REST API (Client and Server implementations)
  • Flask web applications for computer vision services

2. OpenCV

Contains notebooks demonstrating fundamental to advanced OpenCV techniques:

• Image Processing Fundamentals

  • Loading, displaying, and saving images
  • Color filtering and grayscaling
  • Arithmetic and bitwise operations
  • Transformations, translations, and rotations

• Feature Detection and Analysis

  • Contour detection and analysis
  • Edge and corner detection
  • Face and eye detection with Haar cascades
  • Facial landmarks and recognition with Dlib

• Video Processing

  • Webcam access and video capture
  • Video streaming (RTSP, IP cameras)
  • Motion tracking with optical flow
  • Object tracking by color

• Advanced Techniques

  • Background/foreground subtraction
  • Perspective transforms
  • Watershed algorithm for segmentation
  • GrabCut for background removal

Frameworks and Libraries

The notebooks in this repository use various frameworks and libraries, including:

• PyTorch: For deep learning models and neural networks
• Keras/TensorFlow: Alternative deep learning implementations
• OpenCV: For computer vision algorithms and image processing
• Dlib: For facial landmark detection and face recognition
• Flask: For creating web services and applications

Getting Started

1. Clone this repository
2. Install the required dependencies (consider using a virtual environment)
3. Navigate to the notebook of interest and run it in Jupyter

Requirements

The notebooks require Python 3.6+ and the following main packages:

• pytorch
• tensorflow
• keras
• opencv-python
• numpy
• matplotlib
• dlib
• flask

A detailed requirements.txt file will be provided soon.

Contributing

Feel free to contribute to this repository by adding new notebooks, improving existing ones, or fixing issues. Please make sure to follow the existing structure and coding style.

License

This repository contains notebooks and code examples created while learning from educational content.

This collection is shared for educational purposes only. Please use these materials responsibly and give credit to the original instructors/content creators if you reference or build upon this work.

Acknowledgments

Special thanks to all the open-source communities behind the libraries and frameworks used in these notebooks.