Scholarlink

This study discusses the suitability of several pre-trained CNN models for Arabic handwritten character recognition. In particular, this work compares six well-known models: VGG16, VGG19, ResNet50, InceptionV3, MobileNetV2, and MobileNetV3 Large and Small. Accuracy, loss, and training time assessment metrics of the models are evaluated using the Arabic Handwritten Character Dataset (AHCD). This work explores the ability of these models to achieve high classification accuracy using the transfer learning and fine-tuning processes. The results demonstrate significant performance differences, highlighting their advantages and limitations for Arabic character recognition.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Arabic Handwritten Character Recognition: Comparative Performance of Advanced Pre-trained CNN Models

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Othmane Farhaoui,
  • Mohamed Rida Fethi,
  • Ali Omari Alaoui,
  • Imad Zeroual,
  • Ahmad El Allaoui

摘要

This study discusses the suitability of several pre-trained CNN models for Arabic handwritten character recognition. In particular, this work compares six well-known models: VGG16, VGG19, ResNet50, InceptionV3, MobileNetV2, and MobileNetV3 Large and Small. Accuracy, loss, and training time assessment metrics of the models are evaluated using the Arabic Handwritten Character Dataset (AHCD). This work explores the ability of these models to achieve high classification accuracy using the transfer learning and fine-tuning processes. The results demonstrate significant performance differences, highlighting their advantages and limitations for Arabic character recognition.