This README file is the learning structure of the computer vision project, you can learn from any references or source to understand the concept

We suggest you to explore as many as possible references or source to understand the concept well but don't trapped into a rabbit hole. Just learn the basic concept first then go deeper by creating your own hands-on project.

If you are new to programming, I suggested you to use Python as the programming language is easier to understand and have rich ecosystem of libraries for computer vision.

Important: Please don't use my directory in this project because it is randomly created for me playing around with the material I just learned, you must create your own.

Computer Vision learning roadmap (knowledge)

Learning structure :

• What is Computer vision?
• Digital Image Representation and Matrix Operation
• Geometric Operation (translation, rotation, scaling, shearing, etc.)
• Image Filtering
• Histograms and Image Enhancement
• Edge Detection (Sobel, Prewitt, Canny, etc.)
• Image Segmentation
• Feature Detection and Matching
• Object Recognition and Tracking
• Pattern Recognition with Deep Learning
• Data Augmentation and Regulation in Deep Learning for Computer Vision
• Transfer Learning
• Object Detection with Deep Learning (YOLO, SSD, R-CNN, etc.)
• Image Segmentation with Deep Learning (FCN, U-Net, Mask R-CNN, etc.)
• Generative Adversarial Networks (GANs) for Image Generation
• Style Transfer and Image-to-Image Translation
• Reinforcement Learning in Computer Vision
• Multimodal AI (vision + language)
• Explainable AI (XAI) in Computer Vision

Computer Vision learning roadmap (hands-on)

“Learn Computer Vision in 30 Days” by Felipe

a structured 30-day learning plan designed to take a beginner with no prior knowledge of computer vision to a skilled practitioner. The plan consists of 30 diverse computer vision projects spanning various industries, technologies, and application domains. These projects build progressively in complexity and cover foundational concepts, practical implementations, cloud services, and advanced topics.

Overview of the 30-Day Learning Plan

Day	Project/Topic	Key Focus
1	OpenCV Basics	Image handling, blurring, thresholding, edge detection
2	Color Detection	Object detection by isolating colors
3	Face Anonymizer	Face detection and blurring for privacy
4	Text Detection	Benchmarking AWS Textract, EasyOCR, Tesseract
5	Image Classification	Building classifiers using scikit-learn
6	Emotion Detection	Using facial landmarks to classify emotions
7	Feature Extraction	Extracting image features to improve machine learning models
8	Sign Language Detection	Classifying hand landmarks to translate sign language
9	Pneumonia Classification	Chest X-ray analysis and web app development with Streamlit
10	AWS Recognition	Cloud-based object detection without custom model training
11	Object Tracking	YOLO-based real-time object detection and tracking
12	Image Segmentation	Annotating and training models for segmenting images
13	Pose Detection	Building pose detectors for gaming, therapy, and interactions
14	Parking Space Counter	Modular parking spot detection using image segmentation and math
15	Object Detection with YOLOv10	Data preparation and training YOLOv10 models
16	End-to-End Project	Complete project lifecycle from requirements to deliverables
17	Image Processing API	Deploying scalable API on AWS Lambda and API Gateway
18	Dataset Size Analysis	Experimenting with how dataset size affects model performance
19	Document Classification	Real-world freelance project on document categorization
20	Detectron2 Object Detection	Using Facebook AI’s Detectron2 framework
21	Face Recognition with JavaScript	Implementing face recognition using face-api.js
22	Face Attendance System	Real-time face detection and automated attendance recording in Python
23	Intruder Detection	Cloud-based real-time intruder detection with AWS services
24	LangChain Web App	Building chat applications that interact with images
25	Face Generation	Training generative models with Stable Diffusion and DreamBooth
26	Spoofing Detection	Enhancing face attendance system to detect spoofing attempts
27	Face Matching	Comparing faces for identity verification using DeepFace
28	Background Removal Web App	Serverless API for automatic background removal using Segment Anything Model
29	License Plate Detection on Edge	Real-time license plate detection on Raspberry Pi with Coral USB accelerator
30	Synthetic Dataset Generation	Creating synthetic data to overcome real data limitations for training

Key Highlights and Concepts

• Progressive Learning Path: Starts with basic image processing techniques (blurring, edge detection) and advances to complex ML model training, deployment, and cloud integration.
• Diverse Application Areas: Projects span healthcare (pneumonia classification), security (intruder detection, spoofing detection), accessibility (sign language detection), and commercial use cases (e-commerce background removal, attendance systems).
• Hands-on Implementation: Emphasis on building functional applications including web apps, APIs, and real-time systems.
• Use of Popular Tools and Frameworks: OpenCV, YOLO, Detectron2, AWS services, face-api.js, scikit-learn, Streamlit, DeepFace, Stable Diffusion, and LangChain.
• Cloud and Edge Computing: Deployments on AWS Lambda, API Gateway, and Raspberry Pi with hardware accelerators.
• Data-centric Approach: Analysis of dataset size impact and synthetic data generation to improve model robustness.
• Privacy and Security Focus: Including face anonymization, spoofing detection, and intruder alerts to address ethical concerns.
• Real-world Project Insights: Freelance document classification project and end-to-end project walkthrough provide practical exposure beyond tutorials.

Core Insights

• Foundational skills in image processing are crucial for understanding and building complex computer vision systems.
• Machine learning and deep learning models require careful data preparation and feature extraction to achieve high accuracy.
• Cloud services like AWS provide powerful, scalable solutions that simplify deployment and add real-time capabilities.
• Edge computing enables practical deployments in constrained environments like Raspberry Pi for real-time applications.
• Privacy-preserving techniques and security features are integral components of modern computer vision applications.
• Synthetic datasets are valuable tools when real-world data is scarce or difficult to obtain, enhancing model training.