项目作者: AmritK10

项目描述 :
Image classification CNN model on MNIST dataset
高级语言: Jupyter Notebook
项目地址: git://github.com/AmritK10/MNIST-CNN.git
创建时间: 2018-12-13T20:46:14Z
项目社区:https://github.com/AmritK10/MNIST-CNN

开源协议:MIT License

下载


MNIST-CLASSIFICATION

Dataset

The MNIST database (Modified National Institute of Standards and Technology database) is a large database of handwritten digits that is commonly used for training various image processing systems.
The MNIST database of handwritten digits, has a training set of 60,000 28x28 grayscale images of the 10 digits, and a test set of 10,000 examples. It is a subset of a larger set available from NIST. The digits have been size-normalized and centered in a fixed-size image.

Sample images from MNIST test dataset:

github-small

Image classification CNN model on MNIST dataset

The model consists of 2 convolutional layers which are followed by maxpooling layers.The output of these layers is then flattened and fed into the dense layers which give the final output.

Keras implementation accuracy

Train accuracy: 99.66%

Test accuracy: 99.12%

Tensorflow implementation accuracy

Test accuracy: 98.55%