sezan92 / bangladigit Goto Github PK

In this Project I have Developed a Code for Bangla Handwritten Digit Recognizer . Using 2180 Images and SVM with NuSVC() , the Training Accuracy Has been achieved 90% . The Test Folder Contains 30 Images of which 28 Images were predicted accurately.

The Classifier was selected using MLClassifier.py file.

Also Same dataset was used for Deep Convolutional Neural Network using Tensorflow.

Training Accuracy was more than 90% in this case. Although I believe higher accuracy can be achieved by augmenting the data more.

Predicted Images