Computer Vision
Teaching machines to understand images and video: what is in a picture, where it is, and which pixels belong to it.
Your eyes see colors, but your brain sees a dog chasing a ball. Computer vision is about giving computers that second part.
Where it fits
Not shown: 8 more kinds and parts. See the full list below.
+All relationships as a list
Overview
Computer vision tasks differ mainly in how detailed the answer is:
- Image classification: one label for the whole image (“cat”).
- Object detection: a box and a label for every object (“cat here, dog there”).
- Segmentation: a label for every pixel.
The workhorse architecture since AlexNet (2012) has been the CNN, with milestones such as ResNet. Since 2020 the Vision Transformer has shown that attention works for images too.
Because labeled images are expensive, vision relies heavily on data augmentation and on transfer learning: start from a network pretrained on millions of images and fine-tune it on your own small dataset.
Everyday use cases
- Case 01
Your phone camera
Face unlock, portrait-mode blur, searching your photos for "beach" or "dog".
- Case 02
Driver assistance
Spotting lanes, cars, cyclists and traffic signs in real time.
- Case 03
Medicine and industry
Flagging tumors on scans, finding cracks on production lines.