Machine Learning & Computer Vision

History of Computer Vision

Computer Vision: Emulates the human visual system, interpreting images via sensing. Applications include self-driving cars, surveillance, and biomedical imaging.

Classical Era vs. Deep Era

• Classical Era (1960s–2000s): Based on mathematical principles and handcrafted features (e.g., stereovision, projective geometry, SIFT). Rule-based and explicit.
• Deep Learning Era (2012+): Neural network–driven, data-centric, and end-to-end learning.

Computer Vision vs. Machine Learning

• Computer Vision (CV) is a subfield of Machine Learning (ML).
• Features: Measurements for object identification. In classical CV, features are handcrafted; in ML-based CV, models learn them automatically from data.

ML-based CV vs. Classic CV

ML-based CV: Train/test workflow; models learn mappings from data to results.

Classic CV: Predefined feature extraction + simpler classifiers.

Factors Leading to ML Rise

• Improved AI algorithms.
• GPUs enabling massive parallelism (3–4 orders of magnitude faster than CPUs).
• Large datasets (e.g., ImageNet).
• Open-source libraries (Jupyter, TensorFlow, PyTorch).