ML Paper Challenge Day 6 — ImageNet Classification with Deep Convolutional Neural Networks

Papers with Code - ImageNet Classification with Deep Convolutional Neural Networks

Day 6: 2020.04.17
Paper: ImageNet Classification with Deep Convolutional Neural Networks
Category: Model/CNN/Deep Learning/Image Recognition

This is the cornerstone of applying deep CNN in image classification. Before that, the research community thought that such a vision system had to be carefully hand-designed with a deep understanding of the nature of the task. It would never be something like Neural Network by simply providing images and labels.

Approach:

CNN

• “have lots of prior knowledge to compensate for all the data we do not have”
• “capacity can be controlled by varying their depth and breadth, and they also make strong and mostly correct assumptions about the nature of images (namely, stationarity of statistics and locality of pixel dependencies) ”