The ImageNet Large Scale Visual Recognition Challenge (ILSVRC) evaluates algorithms for object detection and image classification at large scale. One high level motivation is to allow researchers to compare progress in detection across a wider variety of objects -- taking advantage of the quite expensive labeling effort. Except 2011 XRCE is SVM based model, CNN-based architecture became dominant from 2012 AlexNet. We will implement LeNet which is earliest CNN architecture proposed by Yann Lecun in 1998. We will use MNIST for demonstrate digit recognition, and I will make some adjustment to enhance performance. By the way,VGG scored 7.3% in 2014. VGG-16, VGG-19 are also famous CNN architecture. Another concept "Inception" is a variation of CNN for better way to stacking layers. For more detail, please check: https://www.kaggle.com/leekltw1/use-tensorboard-s-graph-lenet-to-understand-cnn