Computer Vision - Exam Revision Notes
Mind Map
Detailed Notes
Definition
• Field of AI enabling computers to understand and interpret images & videos like humans.
• Goals: Recognition, measurement, search, interaction with visual data.
• Definitions:
- Ballard & Brown (1982): Build explicit, meaningful descriptions of physical objects from
images.
- Trucco & Verri (1998): Compute 3D properties from one or more images.
- Sockman & Shapiro (2001): Make useful decisions about objects/scenes from images.
Scope
1. Perception & Interpretation – Identify objects, people, scenes, activities.
2. Measurement – Extract 3D world properties.
3. Search & Organization – Manage and retrieve visual content.
Human vs Computer Vision
• Computers: Good at simple, repetitive tasks (with large data).
• Humans: Superior in complex, context-based tasks.
• AI is improving rapidly, changing what is considered 'hard'.
Why is Vision Hard?
1. Incomplete Information – Unknown camera settings, lighting, object shape.
2. Inverse Problem – Need to infer real-world scene from 2D data.
3. High-Dimensional Data – Computationally expensive.
4. No complete model of human visual system.
5. Context dependency – Deciding what is important in an image.
Importance
• Visual cortex occupies ~50% of Macaque brain; ~1/3 of human brain dedicated to vision.
• Billions of images/videos captured daily (Flickr, Facebook, Instagram, YouTube).
Applications
• Consumer Electronics: QR codes, panorama, night mode, FaceID.
• Healthcare: Medical imaging, diagnosis, surgery planning.
• Security: Biometric authentication, CCTV monitoring.
• Industry: Robotics, automated inspection, OCR, number plate recognition.
• Transportation: Self-driving cars, driver monitoring.
• Space: Mars rovers, telescopes.
• Entertainment: AR/VR, film VFX, virtual sports replay.
• Retail: Cashier-less checkout, theft detection.
• Creative: Image/video generation (GANs), photo editing.
Research Areas
• Reconstruction: 3D/4D scene recovery from photos or depth cameras.
• Recognition: Object detection, face recognition, classification.
• Generation: Image-to-image translation, GANs.
• Visual Search: Content-based retrieval.
Why Study Computer Vision?
• Huge data availability from internet and sensors.
• Broad application across multiple domains.
• Critical in automation, AI, and human-computer interaction.
