top of page

Build it up

You are kindly requested to revise your manuscript and submit the updated version to PFDM 2025 before 15-06-2025.

 

Below, you will find all relevant review comments from:

​

  • Scientific Committee,

  • Track Leaders,

  • The Editorial Team
     

Please consider these carefully in preparing your revised manuscript.Once your revisions are complete, you may submit the updated version using the submission link provided at the bottom of this page.We appreciate your contributions and look forward to receiving your revised manuscript.

Enhancing Pavement Crack Detection with Attention-Augmented U-Net Models: A Comparative Study

Reviewer's feedback:

There are not specific remarks about contents. Be carefull that title of 3.2 section should be "Neural" and not "Neutral". "This paper presents an attention-augmented U-Net model for pavement crack detection. The study evaluates the effectiveness of integrating channel and spatial attention mechanisms into pre-trained CNN backbones (VGG16, ResNet50) within the U-Net architecture. Performance is assessed on the Crack500 dataset using standard metrics (Precision, Recall, F1-Score, IOU). 

• The methodology is technically sound. • Dataset and evaluation metrics are appropriate. 


• Comparative analysis of baseline and attention-augmented models is valuable. Ø Table 2 should be captioned as figure Ø Lack of qualitative results (sample crack detection images). Ø No detailed discussion of why “while the introduction of attention mechanisms slightly reduces precision and recall, it enhances overall feature learning and localization.” Ø Proofread and revise for clarity and conciseness. Ø Conclusions must be more precise and provide a better explanation of why the objective of the study was accomplished " 

  • The paper needs an abstract.

Editorial Decision for Conference Proceedings:

It is recommended that authors follow the template, as it already provides a clear structure. Suitable for conference publication.

Track Leader’s Comments (if any):

Please note that some of the track leader’s comments are intended as feedback for future improvements

  • This study aims to enhance crack detection in highway pavements by integrating an attention mechanism within a CNN encoder-decoder, specifically the U-net model.

  • There is no Abstract for the paper. This should be added

  • The authors illustrate the value of incorporating attention mechanisms in enhancing model performance in complex crack detection scenarios using attention-augmented Unet models (A-VGG, A-ResNet). This is reported on quite methodically in this short paper.

  • A big positive is that the authors show a marked improvement in the neural network’s ability to accurately identify and characterize each type of crack. Enhancing the precision and reliability of crack detection in asphalt pavements is very good for practice.

bottom of page