Page 156 - Kaleidoscope Academic Conference Proceedings 2024

2024 ITU Kaleidoscope Academic Conference

Fig. 6. 26X26 Confusion Matrix on Real-time Test Data (Alphabets)

Fig. 7. Numbers Corresponding to their Respective Alphabets in Fig. 4.
Fig. 5. Deep Learning Powered Canvas for OCR

and inclusive education, but also signifies a pivotal step towards
B. Software Requirements transforming learning on a global scale. A comprehensive fea-
1) Front-end: sibility analysis robustly substantiates the technical and market
• HTML5 viability of Alpha-Bit, positioning it as a feasible and impactful
• CSS3 solution. The meticulously planned project timeline foresees the
• JavaScript ES6 development and launch of the Alpha-Bit app within a concise
three-month period. In light of its transformative capabilities,
2) Back-end:
Alpha-Bit emerges as a potent force with the potential to rev-
• Python 3.6 or later
olutionize learning practices worldwide, ushering in a new era of
• Flask web framework
accessible and dynamic educational experiences.
• Required Python libraries (e.g., NumPy, OpenCV for image
processing, TensorFlow or PyTorch for CNN models) REFERENCES
3) OCR Libraries: [1] Y. LeCun et al., ”Backpropagation Applied to Handwritten Zip Code
• OCR libraries for implementing CNN and Sequential models Recognition,” Neural Computation, 1989.
[2] A. Graves et al., ”A novel connectionist system for unconstrained hand-
(e.g., TensorFlow, PyTorch)
writing recognition,” IEEE Transactions on Pattern Analysis and Machine
Notably, deep learning OCR outperforms traditional ap- Intelligence, vol. 31, no. 5, 2009.
[3] T. Qiao et al., ”Seed: Semantics enhanced encoder-decoder framework for
proaches, offering higher accuracy, particularly with handwrit-
scene text recognition,” in Proceedings of the IEEE/CVF Conference on
ten input. In comparison to older OCR techniques relying on Computer Vision and Pattern Recognition, 2020, pp. 13528-13537.
handengineered feature extraction, deep learning models au- [4] Z. Wan et al., ”Tencent ML-Images: A Large-Scale Multi-Label Image
Database for Visual Representation Learning,” IEEE Access, vol. 7, pp.
tonomously learn optimal features from training data, accommo-
172683-172693, 2019.
dating variations in handwriting crucial for recognizing children’s [5] N. Sharma et al., ”Multilingual OCR for Resource-Scarce Languages using
work. Achieving over 99% accuracy on handwritten documents, Deep Learning,” in 2019 International Conference on Document Analysis
deep learning OCR emerges as a reliable tool for assessing and Recognition (ICDAR), 2019, pp. 1055-1062.
[6] J. Su and W. Yang, “Artificial intelligence in early childhood education: A
children’s learning (Reddy & Suruliandi, 2020). The strategic scoping review,” Computers & Education, vol. 160, p. 104025, 2020.
utilization of smartphones, leveraging their ubiquity, intuitive in- [7] S. Papadakis, M. Kalogiannakis, and N. Zaranis, “Teaching mathematics
terface, and AI capabilities such as deep learning OCR, holds the with mobile devices and the Realistic Mathematical Education (RME)
approach in kindergarten,” International Journal of Mobile Learning and
potential to significantly enhance the effectiveness of early AI/CS Organisation, vol. 13, no. 2, pp. 255-277, 2019.
education. By enabling personalized, interactive learning expe- [8] J. C. Wang, C. Y. Hsieh, and S. H. Kung, “The impact of smartphone
riences to extend beyond traditional classrooms and providing use on learning effectiveness: A case study of primary school students,” in
Proceedings of the International Conference on E-Learning in the Workplace
educational apps the capability to ”see” and respond to children’s (ICELW), New York, NY, USA, 2018, pp. 187–195.
handwritten work, smartphones emerge as a developmentally
appropriate medium for introducing foundational AI/CS concepts
to young learners. While acknowledging the need for further
research, this inference underscores the promise of smartphones
as a transformative tool in early childhood education.

VII. CONCLUSION
In conclusion, this research paper introduced Alpha-Bit, a
groundbreaking mobile application that harnesses deep learning-
based Optical Character Recognition (OCR) to unlock new
dimensions in education. By seamlessly converting the digital
canvas into editable text through a smartphone camera, Alpha-
Bit facilitates instantaneous access to adaptable learning content.
This innovative approach not only aligns with the objectives of
Sustainable Development Goal 4 (SDG 4), emphasizing quality
– 112 –