Computer Vision and Image Processing:

---

📌 Project Overview

This project demonstrates key concepts in Computer Vision, ranging from classical image processing (Edge Detection) to feature matching (SIFT/ORB) and modern Deep Learning classification (CNN vs. SVM).

---

📂 Project Tasks

Task 1: Canny Edge Detection

Implementation of the multi-stage Canny algorithm to identify structural boundaries.

• Process: Noise Reduction (Gaussian Blur) → Gradient Calculation (Sobel) → Non-Maximum Suppression → Hysteresis Thresholding.

• Key Feature: Demonstration of threshold adjustments to optimize edge-map clarity.

• Sample:

Task 2: Feature Matching (SIFT & ORB)

Comparison of local feature identification and matching.

• SIFT (Scale-Invariant Feature Transform): High-precision gradient-based matching.

• Sample for SIFT algorithm

• ORB (Oriented FAST and Rotated BRIEF): High-speed, efficient alternative to SURF/SIFT.

• Sample for SIFT algorithm

• Metric: Uses Brute-Force (BF) Matcher with $L2$ Norm for SIFT and Hamming Distance for ORB.

Task 3: Image Classification (Cat vs. Dog)

A benchmarking study between Classical Machine Learning and Deep Learning.

• Classical Model: SVM (Support Vector Machine) with RBF Kernel and flattened pixel features.
• Deep Learning Model: CNN (Convolutional Neural Network) using spatial hierarchies.
• Optimizations: Adam Optimizer, Dropout for regularization, and Pixel Normalization.
• Sample

---

🛠️ Technologies & Libraries

• Python 3.x
• OpenCV: Image processing and feature detection.
• TensorFlow/Keras: Deep Learning model architecture.
• Scikit-Learn: SVM implementation and performance metrics.
• Matplotlib & Seaborn: Results visualization and confusion matrices.

---

🚀 How to Run (Google Colab)

1. Environment Setup: Ensure you have a GPU runtime enabled for Task 3 (Runtime > Change runtime type > GPU).

2. Directory Structure: Create the following folders in the sidebar:

   dataset/
   ├── Cat/
   └── Dog/