From Research to Document Recognition

LeNet systems combined convolution, subsampling, and gradient-based training for handwritten characters and documents. Their deployment-oriented development demonstrated that neural networks could solve practical recognition tasks at scale.