- Hubel and Weisel, 1959, neural response of cats to the image patterns
- Larry Robrets, 1963, get photographic information and features to understand pictures
- Seymour Papert, 1966, summer computer vision project. construction of a visual system.
- David Marr, 1970s, stages of visual reprensation
- Brooks and Binford, 1979, Recognition via parts and human as rigid parts
- Canny and Lowe, 1987, Recognition via edge detection
- 1990s, Shi and Malik, Recognition via grouping
- 1999, SIFT, David Lowe, Recognition via Matching & design of invariant robust -feature vectors
- 2001 Face detection, Viola and Jones, 1st successful applications of machine learning to vision & 1st commercialization of CV
- 2007 PASCAL Visual Object Challenge
- 2009 IMAGENET, Deng, 1.4M images, 1k labels - crowdsourcing to build the dataset & Benchmark competition ○ Deep learning methods have significantly increased the performance ○ DL outperforms humans on the dataset
- 2012 Breakthrough: Alexnet, NeurIPS
- 1958 Perceptrons: One of the earliest algrithms that could learn from data, implemented as a piece of harware, whose weights are stored in potentiometers.
- 1969 Perceptrons cannot learn the XOR function. Minsky and Papert (People didn't realize that muti-layer perceptrons can)
- 1980 Neocognition, Computation model the visual system, inspired by hierarchy of complex and simple cells. No efficient methods to train the model.
- 1986 Backpropagation for computing gradients in neural networks.
- 1998 Convolutional Networks: LeCun et al, 1998, deployed commercially to process handwriting, e.g. number recognitions (algorithm similar to AlexNet)
- 2000 Deep Learning, Hinton and Salakhutdinov: Deep - multi-layer
- 2012 AlexNet trained on ImageNet
- 2012-Present ConvNets are everywhere
- 2015 YOLO
- 2016 AlphaGo
- 2017 Transformer
- 2017 AlphaFold
- 2018 Turing Award: Yoshua Bengio, Geoffrey Hinton, Yann LeCun
- 2022 Stable Diffusion
- Computer Vision still has a long way to go…
- Context aware & Semantic aware
- Physics+
- Algorithm
- Massive Dataset
- Computing Power