History of CV
- Hubel and Wiesel 1959: cells in hirearchical system, receptors
- Larry Roberts, 1963: can we understand the shave and edges of a picture
- How we understand the image
- 从 3D 世界通过 2D 图像来理解 ill-posed problem multiple eyes
- language is 1D sequential 我们可以模拟，但是图像是完全不同的问题
- Recognition via Parts (1970s)
- Edge Detection (1980s) as digital image appeared
- AI winter
- human ability in reconizing objects, scene, faces, etc
- Object Recognition
- Face Recognition
- Internet can proliferate data, booming the study
- artificial neuron
- handmade neuron layers
- backpropagation
- convolutional neural network
- ImageNet Challenge
- AlexNet
- deep learning explosion
- AI global warming period
Porblems:
- harmful stereotypes
- affect people’s lives
- save lives
Deep Learning Basics
image classification
linear classifier: find the hyperplane that seperate different groups.
how to model complex patterns.
nereual networks to model non-linear function
Tasks of CV
- classification
- semantic segmentation
- object detection
- instance segmentation
- video classification
- multimodal video understanding
- visualization
methods:
- CNN
- Self-supervised learning
- Generative modeling
- vision language model
- 3d vision
Glossary
- Cambrian Explosion
- pinhole
- obscura
- apparatus
- overoptimistic
- digress
- seminal
- photon
- wetware
- proliferate
- engineering feat 工程壮举
- watershed moment
- fanfare
- abysmal
- nuanced
- podium
- nuts and bults