Skip to content

Latest commit

 

History

History
14 lines (9 loc) · 943 Bytes

README.md

File metadata and controls

14 lines (9 loc) · 943 Bytes

MNIST-Recognition

Experimenting with basics of Machine Learning and Computer Vision. Trained a model to do basic MNIST Digit and Alphabet Recognition model

Used sigmoid function as a activation function, and hidden layer of size 32 to detect digits 0 to 9, with a accuracy of about 94% and capital letters, with a accuracy of about 91%. In Alphabet model, used batch size of 32 to speeden the training process.

Implemented a 3 layer neural network(2 hidden layers) to improve the accuracy for Alphabet Recognition, pushing it to above 96%

Implement a tensorflow based CNN model to push the accuracy above 98%

Resources I reffered:

The zip file for A_Z training data