A-Build and train a custom CNN network for AHCR. To build a custom CNN, plot the following1. Training loss vs. epoch.2. Validation loss vs. epoch.3. Training accuracy vs. epoch.4. Testing accuracy vs. epoch.References:● Alsayed, Alhag & Li, Chunlin & Ahamed, & Hazim, Mohammed & Obied, Zainab. (2023). ArabicHandwritten Character Recognition Using Convolutional Neural Networks.10.21203/rs.3.rs-3141935/v1.● Alwagdani, M.S.; Jaha, E.S. Deep Learning-Based Child Handwritten Arabic Character Recognitionand Handwriting Discrimination. Sensors 2023, 23, 6774. https://doi.org/10.3390/s23156774.● Altwaijry, N., Al-Turaiki, I. Arabic handwriting recognition system using convolutional neuralnetwork. Neural Comput & Applic 33, 2249–2261 (2021).● Balaha, H.M., Ali, H.A., Youssef, E.K. et al. Recognizing arabic handwritten characters using deeplearning and genetic algorithms. Multimed Tools Appl 80, 32473–32509 (2021).● Arabic Handwritten Character Recognition based on Convolution Neural Networks and SupportVector Machine. (IJACSA) International Journal of Advanced Computer Science and Applications,Vol. 11, No. 8, 2020● Others…B: Retrain the network selected from part A after doing data augmentation.Data augmentation is a powerful technique for enhancing the diversity and the size of your training datawithout the need for additional data collection. By applying various transformations to existing datapoints, you can train models that generalize better to unseen examples and improve their overallperformance.Data augmentation encompasses a wide range of techniques, only a subset of which are suitable fortraining the AHCR system. In this part, you must select at least three data augmentation techniques thatare appropriate and for the model you trained, you need to plot the following:1. Training loss vs. epoch.2. Validation loss vs. epoch.3. Training accuracy vs. epoch.4. Testing accuracy vs. epoch.Compare the results you obtained with the results of part A.C: Select a CNN network from a list of well-known and published CNN architectures, such as LeNet,AlexNet, ResNet, and so on. You must make a tradeoff between accuracy and network complexity withrespect to the problem and the dataset provided. Train it using the data augmentation techniques youused in part B.For the model you trained, you need to plot the following:1. Training loss vs. epoch.2. Validation loss vs. epoch.3. Training accuracy vs. epoch.4. Testing accuracy vs. epoch.Compare the results you obtained with the results of A and N.D: Use a pre-trained CNN network on similar tasks and choose the appropriate transfer learningmethod to fine tune the pretrained network on the given dataset.For the model you trained, you need to plot the following:1. Training loss vs. epoch.2. Validation loss vs. epoch.3. Training accuracy vs. epoch.4. Testing accuracy vs. epoch.Compare the results you obtained with the results of A,B,C. Data setUse the following dataset to train and test the models:https://drive.google.com/file/d/1w1F-BqyY_1KNS1GQhOJHBsIlR0t3nPNu/view?usp=sharinghttps://drive.google.com/file/d/1
A-Build and train a custom CNN network for AHCR. To build a custom CNN, plot the following
1. Training loss vs. epoch.
2. Validation loss vs. epoch.
3. Training accuracy vs. epoch.
4. Testing accuracy vs. epoch.
References:
● Alsayed, Alhag & Li, Chunlin & Ahamed, & Hazim, Mohammed & Obied, Zainab. (2023). Arabic
Handwritten Character Recognition Using Convolutional Neural Networks.
10.21203/rs.3.rs-3141935/v1.
● Alwagdani, M.S.; Jaha, E.S. Deep Learning-Based Child Handwritten Arabic Character Recognition
and Handwriting Discrimination. Sensors 2023, 23, 6774. https://doi.org/10.3390/s23156774.
● Altwaijry, N., Al-Turaiki, I. Arabic handwriting recognition system using convolutional neural
network. Neural Comput & Applic 33, 2249–2261 (2021).
● Balaha, H.M., Ali, H.A., Youssef, E.K. et al. Recognizing arabic handwritten characters using deep
learning and genetic algorithms. Multimed Tools Appl 80, 32473–32509 (2021).
● Arabic Handwritten Character Recognition based on Convolution Neural Networks and Support
Vector Machine. (IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 11, No. 8, 2020
● Others…
B: Retrain the network selected from part A after doing data augmentation.
Data augmentation is a powerful technique for enhancing the diversity and the size of your training data
without the need for additional data collection. By applying various transformations to existing data
points, you can train models that generalize better to unseen examples and improve their overall
performance.
Data augmentation encompasses a wide range of techniques, only a subset of which are suitable for
training the AHCR system. In this part, you must select at least three data augmentation techniques that
are appropriate and for the model you trained, you need to plot the following:
1. Training loss vs. epoch.
2. Validation loss vs. epoch.
3. Training accuracy vs. epoch.
4. Testing accuracy vs. epoch.
Compare the results you obtained with the results of part A.
C: Select a CNN network from a list of well-known and published CNN architectures, such as LeNet,
AlexNet, ResNet, and so on. You must make a tradeoff between accuracy and network complexity with
respect to the problem and the dataset provided. Train it using the data augmentation techniques you
used in part B.
For the model you trained, you need to plot the following:
1. Training loss vs. epoch.
2. Validation loss vs. epoch.
3. Training accuracy vs. epoch.
4. Testing accuracy vs. epoch.
Compare the results you obtained with the results of A and N.
D: Use a pre-trained CNN network on similar tasks and choose the appropriate transfer learning
method to fine tune the pretrained network on the given dataset.
For the model you trained, you need to plot the following:
1. Training loss vs. epoch.
2. Validation loss vs. epoch.
3. Training accuracy vs. epoch.
4. Testing accuracy vs. epoch.
Compare the results you obtained with the results of A,B,C.
Data set
Use the following dataset to train and test the models:
https://drive.google.com/file/d/1w1F-BqyY_1KNS1GQhOJHBsIlR0t3nPNu/view?usp=sharinghttps://drive.google.com/file/d/1
Step by step
Solved in 2 steps
