Character_README.md

Link of the trained model

• link of the jupyter notebook here

Training and validation accuracy and loss

Example of prediction of character

Confusion Matrix For character

Evaluation and Training accuracies For character

• Evaluation Accuracy: 0.9953

• Training Accuracy: 0.9997

Accuracy, Precision, Recall, F1 Score for all class For character

Overall	Accuracy	Precision	Recall	F1 Score
All	0.9491	0.9533	0.9491	0.9491

Accuracy, Precision, Recall, F1 Score for different class For character

Class	True Samples	Classified Samples	Accuracy	Precision	Recall	F1 Score
0	1129	1146	97.0771%	95.637%	97.0771%	96.3516%
1	1128	1113	98.4929%	99.8203%	98.4929%	99.1522%
2	1129	1104	96.9885%	99.1848%	96.9885%	98.0743%
3	1129	1151	98.8485%	96.9592%	98.8485%	97.8947%
4	1128	1490	97.6064%	73.8926%	97.6064%	84.1100%
5	1129	1154	96.2799%	94.1941%	96.2799%	95.2256%
6	1129	1124	96.1027%	96.5302%	96.1027%	96.3160%
7	1129	1292	95.8370%	83.7461%	95.8370%	89.3845%
8	1129	1127	97.2542%	97.4268%	97.2542%	97.3404%
9	1129	1210	96.8999%	90.4132%	96.8999%	93.5442%
10	1129	1248	99.0257%	89.5833%	99.0257%	94.0682%
11	1129	1153	99.1143%	97.0512%	99.1143%	98.0719%
12	1059	1081	96.8839%	94.9121%	96.8839%	95.8879%
13	1058	1117	98.1096%	92.9275%	98.1096%	95.4483%
14	1059	1109	96.7894%	92.4256%	96.7894%	94.5572%
15	1059	1068	99.43%	98.60%	99.43%	99.01%
16	1059	1051	98.30%	99.05%	98.30%	98.67%
17	1059	1044	97.92%	99.33%	97.92%	98.62%
18	1059	1049	95.28%	96.19%	95.28%	95.73%
19	1059	1135	96.41%	89.96%	96.41%	93.07%
20	1059	1071	95.66%	94.58%	95.66%	95.12%
21	1059	1190	97.73%	86.97%	97.73%	92.04%
22	1059	1091	98.49%	95.60%	98.49%	97.02%
23	1099	993	89.54%	99.09%	89.54%	94.07%
24	1099	963	86.44%	98.65%	86.44%	92.14%
25	1099	918	82.89%	99.24%	82.89%	90.33%
26	1099	782	69.88%	98.21%	69.88%	81.66%
27	1099	1092	98.09%	98.72%	98.09%	98.40%
28	1099	1076	97.73%	99.81%	97.73%	98.76%
29	1099	1000	89.63%	98.50%	89.63%	93.85%
30	1099	1077	97.63%	99.63%	97.63%	98.62%
31	1098	1051	93.81%	98.00%	93.81%	95.86%
32	1099	978	85.44%	96.01%	85.44%	90.42%
33	1099	1042	93.45%	98.56%	93.45%	95.94%
34	1099	1091	97.27%	97.98%	97.27%	97.63%

Model Summery

Layer (type)	Output Shape	Param #
conv2d	(None, 26, 26, 64)	640
activation	(None, 26, 26, 64)	0
max_pooling2d	(None, 25, 25, 64)	0
batch_normalization	(None, 25, 25, 64)	256
conv2d_1	(None, 23, 23, 128)	73856
activation_1	(None, 23, 23, 128)	0
max_pooling2d_1	(None, 22, 22, 128)	0
batch_normalization_1	(None, 22, 22, 128)	512
conv2d_2	(None, 22, 22, 192)	24768
activation_2	(None, 22, 22, 192)	0
batch_normalization_2	(None, 22, 22, 192)	768
conv2d_3	(None, 20, 20, 192)	331968
activation_3	(None, 20, 20, 192)	0
batch_normalization_3	(None, 20, 20, 192)	768
conv2d_4	(None, 18, 18, 128)	221312
activation_4	(None, 18, 18, 128)	0
max_pooling2d_2	(None, 17, 17, 128)	0
batch_normalization_4	(None, 17, 17, 128)	512
flatten	(None, 36992)	0
dense	(None, 2048)	75761664
activation_5	(None, 2048)	0
dropout	(None, 2048)	0
batch_normalization_5	(None, 2048)	8192
dense_1	(None, 2048)	4196352
activation_6	(None, 2048)	0
dropout_1	(None, 2048)	0
batch_normalization_6	(None, 2048)	8192
dense_2	(None, 800)	1639200
activation_7	(None, 800)	0
dropout_2	(None, 800)	0
batch_normalization_7	(None, 800)	3200
dense_3	(None, 35)	28035
activation_8	(None, 35)	0
Total params	82,300,195
Trainable params	82,288,995
Non-trainable params	11,200