Grad-Cam Visualization Of Arabic Letter Character Prediction DOI

Asroni Asroni,

Cahya Damarjati,

Dhimas Rizki Akbar

et al.

Published: Dec. 1, 2023

Arabic letters, commonly called hijaiyah present a considerable challenge in acquisition and mastery. Introducing letters is significant subject due to the inherent challenges associated with their composition. This study aims compare class activation visualization characters by employing custom model contrasting it widely used models, namely AlexNet LeNet. The employed utilizes Class Activation Mapping (CAM) technique demonstrate its understanding of character identification process effectively. approach facilitates observation key focal points when identifies certain character. identify elements that contribute effectiveness Convolutional Neural Network (CNN) accurately recognizing characters. will be achieved training CNN using substantial dataset specifically emphasizes recognition. employ visualize results. results this not only offer comprehensive comprehension model's detection. However, they also assist identifying any problems may arise during procedure. outcomes research would enhance capacity script, hence facilitating implementation assistance for handling text damaged or blurred. In investigation, was observed performance surpassed LeNet convolutional neural network models. Training on consisting 13,440 data points, notable accuracy rate 97.38%. Additionally, exhibited loss 9.07% at epoch 50. interim, demonstrated 96.15% 93.12%, losses 15.88% 21.90%.

Language: Английский

Adaptive and Explainable Deep Learning-Based Rapid Identification of Architectural Cracks DOI Creative Commons

Jiang-Yi Luo,

Yucheng Liu

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 111741 - 111751

Published: Jan. 1, 2024

Language: Английский

Citations

0
Investigation of a Neural Network for Dolphin Whistle Detection Through Heatmaps DOI

Jurica Jerinic,

Alberto Testolin, Roee Diamant

et al.

Published: Oct. 28, 2024

Language: Английский

Citations

0
Remote sensing-based scene classification by feature fusion and extraction with ensemble classifier employing machine learning approaches DOI

A. Arulmurugan,

R. Kaviarasan,

Parimala Garnepudi

et al.

Journal of Intelligent & Fuzzy Systems, Journal Year: 2023, Volume and Issue: unknown, P. 1 - 13

Published: Nov. 27, 2023

This research focuses on scene segmentation in remotely sensed images within the field of Remote Sensing Image Scene Understanding (RSISU). Leveraging recent advancements Deep Learning (DL), particularly Residual Neural Networks (RESNET-50 and RESNET-101), proposes a methodology involving feature fusing, extraction, classification for categorizing remote sensing images. The approach employs dataset from University California Irvine (UCI) comprising twenty-one groups pictures. undergo pre-processing, extraction using mentioned DL frameworks, subsequent categorization through an ensemble structure combining Kernel Extreme Machine (KELM) Support Vector (SVM). paper concludes with optimal results achieved performance comparison analyses.

Language: Английский

Citations

0
Grad-Cam Visualization Of Arabic Letter Character Prediction DOI

Asroni Asroni,

Cahya Damarjati,

Dhimas Rizki Akbar

et al.

Published: Dec. 1, 2023

Arabic letters, commonly called hijaiyah present a considerable challenge in acquisition and mastery. Introducing letters is significant subject due to the inherent challenges associated with their composition. This study aims compare class activation visualization characters by employing custom model contrasting it widely used models, namely AlexNet LeNet. The employed utilizes Class Activation Mapping (CAM) technique demonstrate its understanding of character identification process effectively. approach facilitates observation key focal points when identifies certain character. identify elements that contribute effectiveness Convolutional Neural Network (CNN) accurately recognizing characters. will be achieved training CNN using substantial dataset specifically emphasizes recognition. employ visualize results. results this not only offer comprehensive comprehension model's detection. However, they also assist identifying any problems may arise during procedure. outcomes research would enhance capacity script, hence facilitating implementation assistance for handling text damaged or blurred. In investigation, was observed performance surpassed LeNet convolutional neural network models. Training on consisting 13,440 data points, notable accuracy rate 97.38%. Additionally, exhibited loss 9.07% at epoch 50. interim, demonstrated 96.15% 93.12%, losses 15.88% 21.90%.

Language: Английский

Citations

0