OCR-Malayalam

Classification

	Model	Parameters	Feature Extraction	No. of classes	No of images / Class	Accuracy	Validation Accuracy
1	SVM	kernel='rbf' , gamma=0.1, C=1000	None	44	1000		96%
2	SVM	kernel=rbf c=100 gamma=0.02	PCA, Sobel, Robert, HOG	44	2088		99%
3	CNN	5 Conv, Softmax Classifier	None	44	1000	98.80%	99.87
4	CNN	5 Conv, Softmax Classifier	None	122	1000	96.53%	97.05%
5	ResNet	51 Layers	None	122	4116	99.61%	98.69%
6	SVM	8 Layers SVM Classifier	Alexnet Layers	122	4116	99.41%	97.15%
7	RandomForest	gini , 100 estimators	None	36	3060	95.00%	94.79%
8	RandomForest	gini , 100 estimators	None	44	2088	96.00%	94.99 %