Information / AI & Robotics

Convolutional networks (LeNet)

Yann LeCun's convolutional networks, trained by backpropagation (1989), read handwritten digits and became the blueprint for image AI.

Inspired by Hubel and Wiesel's visual cortex and by Kunihiko Fukushima's Neocognitron (1980), Yann LeCun and colleagues at Bell Labs trained a convolutional network by backpropagation to read handwritten ZIP-code digits in 1989. Its successor LeNet-5 (1998) was used commercially to read bank cheques. Held back by small data sets and slow computers, the design only took over image recognition in 2012.