Definition
Computer vision covers classification, object detection, segmentation, OCR, pose estimation, and video understanding. Deep CNNs and vision transformers power most modern systems.
It pairs with language in multimodal models that caption images or answer visual questions.
In simple terms
Computer vision is teaching cameras to describe what they see — not just record pixels, but name objects and relationships.
Where you see it
- Face unlock and photo search.
- Warehouse robotics and quality inspection.
- Medical imaging triage.
How it works
1.Capture pixels
Images or video frames as input tensors.
2.Extract features
CNN or ViT encodes visual patterns.
3.Predict
Labels, boxes, masks, or captions.
Why it matters
- Vision is how AI interacts with the physical world of photos, screens, and cameras.
Often confused
Computer vision solved all visual tasks.
Robustness to lighting, rare objects, and adversarial examples remains hard.