Using those aforementioned strategies with our proposed encoder-decoder architecture we are able to achieve new state of the art results in all datasets. The former approach generally yields increased performance results, whereas the latter allows for more reproducible and better representability of a methods results. We introduce a decoder design which can be added to existing feature encoder backbone architectures such as ResNet, VGG or EfficientNet to complete the U-shape of the network. Additionally, we introduce two techniques into the field of surface crack segmentation, previously not used there: Generating results using test-time-augmentation and performing a statistical result analysis over multiple training runs. In this work we propose novel surface crack segmentation methods using an encoder-decoder based deep learning architecture. The performance evaluation of our method is carried out on four publicly available crack segmentation datasets. We also examine the use of different encoder strategies and introduce a data augmentation policy to increase the amount of available training data. Specifically we propose a decoder-part for an encoder-decoder based deep learning architecture for semantic segmentation and study its components to achieve increased performance. In this work we propose optimized deep encoder-decoder methods consisting of a combination of techniques which yield an increase in crack segmentation performance. Surface crack segmentation poses a challenging computer vision task as background, shape, color and size of cracks vary.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |