Skip to content

Repository to support the paper titled "Generative Adversarial Networks for Data Augmentation in Structural Adhesive Inspection" published in MDPI Applied Sciences.

Notifications You must be signed in to change notification settings

RicardoSPeres/GAN_Synth_Adhesive

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

32 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Generative Adversarial Networks for Data Augmentation in Structural Adhesive Inspection

We leverage recent advances regarding GAN training in limited data regimes to generate synthetic images of structural adhesive defects such as the ones seen below. We demonstrate that these realistic synthetic samples can be used to augmented scarce datasets and improve the performance of state-of-the-art object detection models in the automated inspection of such adhesive applications.

Article: https://www.mdpi.com/2076-3417/11/7/3086

Citation

If you use our data in your research or wish to refer to the results published in the paper, please use the following BibTeX entry.

@article{peres2021generative,
  title={Generative Adversarial Networks for Data Augmentation in Structural Adhesive Inspection},
  author={Peres, Ricardo Silva and Azevedo, Miguel and Ara{\'u}jo, Sara Oleiro and Guedes, Magno and Miranda, F{\'a}bio and Barata, Jos{\'e}},
  journal={Applied Sciences},
  volume={11},
  number={7},
  pages={3086},
  year={2021},
  publisher={Multidisciplinary Digital Publishing Institute}
}

Peres, R.S.; Azevedo, M.; Araújo, S.O.; Guedes, M.; Miranda, F.; Barata, J. Generative Adversarial Networks for Data Augmentation in Structural Adhesive Inspection. Appl. Sci. 2021, 11, 3086. https://doi.org/10.3390/app11073086.

Dataset

The dataset used in the paper can be found at: https://github.com/RicardoSPeres/GAN_Synth_Adhesive/releases/latest

Generating synthetic defect images with StyleGAN2-ADA

Visualizing truncation traversals

In this case the seeds used were 3845 and 55832, truncation values between -1.0 and 3.0 with increments of 0.05.

Exploring the latent space with GANSpace

Visualizing the effect of random seeds with different truncation and scale values

Comparing defect detection results with YOLOv4-Tiny models trained on real and augmented datasets

Real Augmented

Back to top

About

Repository to support the paper titled "Generative Adversarial Networks for Data Augmentation in Structural Adhesive Inspection" published in MDPI Applied Sciences.

Resources

Stars

Watchers

Forks

Packages

No packages published