Convolutional neural networks (LeNet)
Redes neuronales convolucionales (LeNet)
- Date
- 1989-1998 · confidence high
- Place
- AT&T Bell Labs, Holmdel, New Jersey, United States
- Credited to
- Yann LeCun, with Léon Bottou, Yoshua Bengio and Patrick Haffner; Kunihiko Fukushima's Neocognitron (1980) as a precursor
- People
- 👤 Kunihiko Fukushima1936–
- Field
- Knowledge & symbols
What it is
Networks with shared filters inspired by the visual cortex, trained with backpropagation to read handwritten digits on cheques. They were the first industrial application of deep learning.
What it led to
- The automatic reading of cheques and postal codes
- Modern computer vision (AlexNet) and facial recognition
- Self-driving cars and AI-assisted image diagnosis
Built on
- Backpropagation for training neural networks1986
- Neocognitron
- Hubel and Wiesel's neurophysiology
Made possible
- Self-driving car (VaMoRs, DARPA Grand Challenge, Waymo)1986-2018
- AlexNet and the deep learning revolution2012
- facial recognition
- AI medical diagnosis