Handwritten Gujarati Numerals Classification Based on Deep Convolution Neural Networks Using Transfer Learning Scenarios

Abstract

In recent years, handwritten numeral classification has achieved remarkable attention in the field of computer vision. Handwritten numbers are difficult to recognize due to the different writing styles of individuals. In a multilingual country like India, negligible research attempts have been carried out for handwritten Gujarati numerals recognition using deep learning techniques compared to the other regional scripts. The Gujarati digit dataset is not available publicly and deep learning requires a large amount of labeled data for the training of the models. If the number of annotated data is not sufficient enough to train Convolutional Neural Networks (CNN) from the scratch, transfer learning can be applied. However, the issue arises by using transfer learning is that how deep to fine-tune the pre-trained convolutional neural network while training the target model. In this paper, we addressed these problems using three deep transfer learning scenarios to classify handwritten Gujarati numerals from the images of zero to nine. We presented transfer learning scenarios using ten pre-trained CNN architectures including LeNet, VGG16, InceptionV3, ResNet50, Xception, ResNet101, MobileNet, MobileNetV2, DenseNet169 and EfficientNetV2S to find the best performing model by freezing and fine-tuning the weight parameters. We implemented the pretrained models using a self-created handwritten Gujarati digit dataset with 8000 images of zero to nine digits with data augmentation. Exhaustive experiments are performed using various performance evaluation matrices. EfficientNetV2S model showed promising results among all the models including three transfer learning scenarios and achieved 98.39% training accuracy, 97.92% testing accuracy, 97.69% f1-score, and 97.15% AUC. Our handwritten Gujarati digit dataset is available on https://github.com/Parth-Goel/gujaratihandwritten-digit-dataset/.

Advancing Multilingual Handwritten Numeral Recognition With Attention-Driven Transfer Learning

“…The study [2] explores the categorization of handwritten Gujarati numerals ranging from zero to nine using deep transfer learning techniques. The investigation utilized ten preexisting CNN architectures, namely LeNet, VGG16, InceptionV3, ResNet50, Xception, ResNet101, MobileNet, MobileNetV2, DenseNet169, and EfficientNetV2S, to identify the most suitable model through the fine-tuning of weight parameters.…”

“…(1) 𝑅𝑒𝑐𝑎𝑙𝑙 = 𝑇𝑟𝑢𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝑇𝑃) 𝑇𝑟𝑢𝑒 𝑃𝑜𝑠𝑖𝑡𝑖𝑣𝑒𝑠 (𝑇𝑃) + 𝐹𝑎𝑙𝑠𝑒 𝑁𝑒𝑔𝑒𝑡𝑖𝑣𝑒𝑠(𝐹𝑃) (2) Recall is the proportion of relevant instances that have been correctly identified. It quantifies the percentage of true positives that were accurately recognized.…”

“…Handwritten numeral recognition plays a crucial role in the fields of image processing and computer vision, making significant contributions to diverse applications. This technology is widely used in a wide range of domains, including bank check processing, postal code recognition, and the analysis of historical handwritten documents [1][2][3][4][5]. Accurate recognition of handwritten numerals is crucial in delicate systems like medical records and biometric identification, where mistakes can lead to serious consequences [6][7][8][9][10].…”

“…For instance, Abhishek Sethy proposed a subsequential method that combined LS-SVM and RF classifiers, achieving an impressive overall accuracy of 99.01% on handwritten Odia character dataset [16]. In recent years, the advent of deep learning has led to a renewed focus on improving the effectiveness of systems for recognizing handwritten scripts, particularly through the use of RNNs (Recurrent Neural Networks) and CNNs (Convolutional Neural Networks) [2,19,20]. Notably, Abu Sufian introduced an end-to-end approach utilizing a densely connected convolutional neural network (BDNet) and achieved a remarkable accuracy of 99.78% on ISI Bengali handwritten numerals [21].…”

Incremental Recognition of Multi-Style Tibetan Character Based on Transfer Learning

“…In existing exploratory research on transfer learning, significant attention has been given to the following directions: 1) the degree of fine-tuning of pre-trained models, 2) few-shot transfer learning, and 3) zero-shot transfer learning. To address the question of how much fine-tuning should be applied to the pre-trained convolutional neural network when training the target model, Goel et al [25] employed three transfer learning strategies to demonstrate the impact of different finetuning strategies on model performance. They discovered that connecting two newly adapted fully connected layers to the final convolutional layer of the pre-trained model and fine-tuning only the fully connected layers improved model performance and significantly reduced training time.…”