Scholarlink

The faster ResNet50 achieved 98% accuracy, and Xception achieved 93%. The objective is to compare these models in terms of the performance of the detection of blade surface defects. Materials and Methods: Optimization of rotor blade surface defect detection will be conducted by Faster ResNet50 and Xception. And, for each model, datasets are divided into two categories, and then 10 samples are chosen carefully in each category. To analyze and prove the performance, a statistical test was conducted using the SPSS tool with alpha = 0.8 and beta = 0.2. The G control value set to 87 was found to be predictive of the dataset. Results: A faster ResNet50 he realized his 98% accuracy that outperformed Xception which only reached to accuracy of 93% while detecting blade surface defects. The SPSS analysis came out with a significant difference between the two models with the p-value at 0.0085 (p < 0.05). This will consequently mean that ResNet50 is statistically much better than Xception in defect detection on the leaf surface.

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

Blade Surface Defect Detection in Images Using Faster ResNet50 and Xception for Better Accuracy

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Chittineni Harshitha Chowdary,
  • C. Rohith Bhat

摘要

The faster ResNet50 achieved 98% accuracy, and Xception achieved 93%. The objective is to compare these models in terms of the performance of the detection of blade surface defects. Materials and Methods: Optimization of rotor blade surface defect detection will be conducted by Faster ResNet50 and Xception. And, for each model, datasets are divided into two categories, and then 10 samples are chosen carefully in each category. To analyze and prove the performance, a statistical test was conducted using the SPSS tool with alpha = 0.8 and beta = 0.2. The G control value set to 87 was found to be predictive of the dataset. Results: A faster ResNet50 he realized his 98% accuracy that outperformed Xception which only reached to accuracy of 93% while detecting blade surface defects. The SPSS analysis came out with a significant difference between the two models with the p-value at 0.0085 (p < 0.05). This will consequently mean that ResNet50 is statistically much better than Xception in defect detection on the leaf surface.