Scholarlink

This study introduces an innovative deep ensemble learning framework aimed at enhancing digital watermark classification, addressing challenges in detecting subtle watermark patterns amidst distortions. Traditional single-architecture methods often fail to capture essential features, so this framework integrates MobileNetV2, ResNet101V2, and DenseNet201 using a weighted voting system based on their performance. DenseNet201 achieved the highest metrics, and the framework attained 94.44% accuracy, demonstrating its effectiveness. The novel weighted voting mechanism dynamically assigns weights to classifiers, enhancing robustness against noise and distortions. This scalable framework has applications in digital forensics and copyright protection.

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

Weighted Voting Based Deep Ensemble Learning Framework for Digital Watermark Classification

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Bharathi Pilar,
  • Safnaz

摘要

This study introduces an innovative deep ensemble learning framework aimed at enhancing digital watermark classification, addressing challenges in detecting subtle watermark patterns amidst distortions. Traditional single-architecture methods often fail to capture essential features, so this framework integrates MobileNetV2, ResNet101V2, and DenseNet201 using a weighted voting system based on their performance. DenseNet201 achieved the highest metrics, and the framework attained 94.44% accuracy, demonstrating its effectiveness. The novel weighted voting mechanism dynamically assigns weights to classifiers, enhancing robustness against noise and distortions. This scalable framework has applications in digital forensics and copyright protection.