Online Handwritten Bangla and Devanagari Character Recognition by using CNN: A Deep Learning Concept

Chakraborty, Rajatsubhra; Mukherjee, Debarati; Bhattacharyya, Ankan; Mukherjee, Himadri; Sur, Monoj Kumar; Sen, Shibaprasad

doi:10.1109/icce50343.2020.9290566

Cited by 4 publications

(1 citation statement)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With regular data, GoogleNet obtains an average recognition accuracy of 93%. Chakraborty, et al [8] In a study, various Deep learning approaches consisting of convolutional neural networks (CNN), deep belief networks (DBN), CNN with abandonment, CNN with loss and Gaussian filters, and CNN with dropout and Gaussian filters for handwritten Bangla digit identification. These methods were tested on the CMATERdb 3.1.1 Bangla numeral image database.…”

Section: Literature Reviewmentioning

confidence: 99%

Bangla Handwritten Character Recognition Using Deep Learning Approaches and its Explainability With AI

2024

IJFMR

View full text Add to dashboard Cite

The realm of Bangla handwritten character recognition (BHCR) has long been overshadowed by the dominance of more mainstream languages, despite Bangla's status as Deep learning (DL) approaches have led to substantial improvements in handwritten character recognition (BHCR) for Bangla, one of the humanity's frequently spoken languages. These methods generally an optimal fit for BHCR because they are good at selecting high-level characteristics from intricate information. In our comprehensive study, we meticulously explored the efficacy of twelve DL models on the arduous task of Bangla character recognition, meticulously evaluating their performance on two distinct datasets: a handwritten character dataset and CMATERDB [1], comprising a formidable collection of 15,000 images. Additionally, we provided an audit of the DL models' achievements for BHCR. Among the compared models are LSTM, Bi-LSTM, CNN, Inception, VGG, and ResNet. we achieved the maximum performance at ResNet152V2. In this study, One of the most exquisite identification rates for Bengali character recognition currently available has been demonstrated by the suggested technique, which displayed an adequate 98.76% recognition accuracy on the dataset.

show abstract

Section: Literature Reviewmentioning

confidence: 99%