Abstract
Text detection and identification are crucial elements in comprehending semantics in a natural scene. As a result of human ideas and manipulation, text in a natural scene carries meaningful information. In the state of Assam, Assamese is the official language and has over 15 million speakers worldwide. But, since no major research or application is being developed on a real-time natural scene Assamese text detection system, it is a driving force for us to develop and contribute the same. The current state-of-the-art Optical Character Recognition (OCR) engine can only detect document-based images and is unable to detect text in natural scene images of different languages, colors, fonts, sizes, etc. Also, traditional text detection methods like Sliding Window produce huge numbers of weights and make them unusable for real-time text detection. Aiming at the problems of scene text detection, we proposed several versions of the YOLO-DarkNet algorithm for achieving real-time detection, and YOLOv5s is trained on discrete GPUs. Several hyper-parameter tuning methods are applied to optimize our proposed model. The hyper-parameter optimizer Stochastic Gradient Descent (SGD) minimizes the extremely high computing power by attaining faster iterations in exchange for a reduced convergence rate. After training, our proposed model based on YOLOv5s optimized with SGD generates a weight of 13.6 MB with 7.01 million parameters, and the performance on custom datasets achieves a detection speed of 13.3 ms and precision (mAP) of 94.3%.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
References
Smith R (2007) An overview of the tesseract OCR engine. In: Ninth International Conference on Document Analysis and Recognition (ICDAR 2007). pp. 629–633. https://doi.org/10.1109/ICDAR.2007.4376991
Brisinello M, Grbic R, Vranješ M, Vranješ D (2019) Review on text detection methods on scene images. In: 61st International Symposium ELMAR-2019. Zadar, Croatia, pp 23–25.
Ye Q, Doermann D (2015) Text detection and recognition in imagery: a survey. IEEE Trans Pattern Anal Mach Intell 37(7):1480–1500
Behzadi M, Safabakhsh R (2018) Text detection in natural scenes using fully convolutional DenseNets. In: 2018 4th Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS). pp 11–14. https://doi.org/10.1109/ICSPIS.2018.8700562
Mehta MA, Pote SA (2018) Text detection from scene videos having blurriness and text of different sizes. 2018 IEEE Punecon 1–4. https://doi.org/10.1109/PUNECON.2018.8745375
Mokayed H, Shivakumara P, Liwicki M, Pal U (2020) A new defect detection method for improving text detection and recognition performances in natural scene images. In: 2020 Swedish Workshop on Data Science (SweDS). pp 1−7. https://doi.org/10.1109/SweDS51247.2020.9275589
He W, Zhang Y, Yin F, Liu CL (2018) Multi-oriented and multi-lingual scene text detection with direct regression. IEEE Trans Image Process 27(11):5406–5419. https://doi.org/10.1109/TIP.2018.2855399
Redmon J, Farhadi A (2018) YOLOv3: an incremental improvement. ArXiv https://arxiv.org/abs/1804.02767
Wang H, Zhang Z (2019) Text detection algorithm based on improved YOLOv3. In: 2019 IEEE 9th International Conference on Electronics Information and Emergency Communication (ICEIEC). pp 147–150. https://doi.org/10.1109/ICEIEC.2019.8784576
Thakare S, Kamble A, Thengne V, Kamble UR (2018) Document segmentation and language translation using tesseract-OCR. In: 2018 IEEE 13th International Conference on Industrial and Information Systems (ICIIS). pp 148–151. https://doi.org/10.1109/ICIINFS.2018.8721372
Deep learning for aircraft wake vortex identification-scientific figure on ResearchGate. https://www.researchgate.net/figure/YOLO-v3-algorithm-flow-chart-A-convolutional-neural-network-usually-consists-of-an-input_fig2_337451395. Accessed 14 May 2022.
Bansal M, Jain G (2017) Improvement of english-hindi machine translation using conceptNet. In: 2017 Recent Developments in Control Automation & Power Engineering (RDCAPE). pp 198−202. https://doi.org/10.1109/RDCAPE.2017.8358266
Deka A, Sarmah P, Samudravijaya K, Prasanna SRM (2019) Development of Assamese text-to-speech system using deep neural network. In: 2019 National Conference on Communications (NCC). pp 1−5. https://doi.org/10.1109/NCC.2019.8732262
Adarsh P, Rathi P, Kumar M (2020) YOLO v3-tiny: object detection and recognition using one stage improved model. In: 2020 6th International Conference on Advanced Computing and Communication Systems (ICACCS). pp. 687–694. https://doi.org/10.1109/ICACCS48705.2020.9074315
Pytorch-YOLOv5. https://pytorch.org/hub/ultralytics_yolov5/. Accessed 12 Dec 2021.
YOLOv5 Github page. https://github.com/ultralytics/yolov5. Accessed 14 May 2022.
YOLOv5 study notes. https://www.codetd.com/en/article/12277758. Accessed 14 May 2022.
Menghan Z, Zitian L, Yuncheng S (2020) Optimization and comparative analysis of YOLOV3 target detection method based on lightweight network structure. In: 2020 IEEE International Conference on Artificial Intelligence and Computer Applications (ICAICA). pp 20−24. https://doi.org/10.1109/ICAICA50127.2020.9182679
Koo J, Klabjan D, Utke J (2020) Combined convolutional and recurrent neural networks for hierarchical classification of images. In: 2020 IEEE International Conference on Big Data (Big Data). pp 1354−1361. https://doi.org/10.1109/BigData50022.2020.9378237
Nie X, Yang M, Liu RW (2019) Deep neural network-based robust ship detection under different weather conditions. In: 2019 IEEE Intelligent Transportation Systems Conference (ITSC). pp. 47−52. https://doi.org/10.1109/ITSC.2019.8917475
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
About this paper
Cite this paper
Enghi, M., Talukdar, A.K., Sarma, K.K. (2023). Real-Time Detection of Natural Scene Assamese Texts Using Deep Learning. In: Bhattacharjee, R., Neog, D.R., Mopuri, K.R., Vipparthi, S.K. (eds) Artificial Intelligence and Data Science Based R&D Interventions. NERC 2022. Springer, Singapore. https://doi.org/10.1007/978-981-99-2609-1_4
Download citation
DOI: https://doi.org/10.1007/978-981-99-2609-1_4
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-99-2608-4
Online ISBN: 978-981-99-2609-1
eBook Packages: Computer ScienceComputer Science (R0)