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.