Main Concept
- Computer Vision focuses on helping machines “see” and interpret visual information.
Computer Vision and AI
Traditional Computer Vision (pre-AI era) systems were based on mathematical principles, signal processing and explicitly programmed logic rather than learning from data.
They relied heavily on:
- Mathematical algorithms for image processing (filtering, edge detection, morphological operations)
- Classical feature extraction methods like SIFT, SURF, or HOG descriptors
- Geometric approaches for tasks like stereo vision and 3D reconstruction
- Statistical pattern recognition techniques
- Hand-crafted rules and heuristics
Modern AI-powered Computer Vision uses machine learning and deep learning to:
- Automatically learn features from data instead of hand-crafting them
- Handle much more complex recognition tasks
- Achieve higher accuracy on challenging problems like object detection in cluttered scenes
NOTE
Today when people say “computer vision,” they’re usually referring to the AI-powered approaches because they’ve become so dominant and effective.
AWS services
AWS computer vision services (like Amazon Rekognition) are primarily AI/ML-based, but the underlying field encompasses both traditional mathematical approaches and modern AI methods.