Tamil Handwritten Character Recogniton Model
Tamil Handwritten character recognition using convolutional neural networks.
Dataset: IWFHR-10
Samples from the dataset
- Let’s see the distribution of training samples to check whether the dataset is balanced or imbalanced.
- We can observe that there is a slight class imbalance (i.e., some classes have more training samples over other classes) problem
Model Architecture:
Training
- Loss: Categorical Cross Entropy
- Optimization: Nesterov Accelarated GD
Performance
- The Performance Measure of the model for training and valiadation split.
Confusion Matrix:
- The number of classes in the problem is 127.THerefore, We can plot the confusion matrix to get a coarse view of the performence of the model.
