Optimized deep encoder-decoder methods for crack segmentation

König, Jacob; Jenkins, M.; Mannion, Mike; Barrie, Peter; Morison, Gordon

doi:10.1016/j.dsp.2020.102907

Cited by 62 publications

(60 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Chen et al [7] develop a rotation-invariant fully convolution network (FCN) called ARF-Crack to utilize the rotation-invariant property of cracks. Konig et al [9] introduce a new design for the decoder, which leads to improved performance. While these methods demonstrate promising results, it is important to note that they allow a tolerance margin of a few pixels between predictions and ground truth during evaluation.…”

Section: Ground Truthmentioning

confidence: 99%

“…DRIVE has 40 color fundus photographs. For training and test sets, we split AigleRN equally into 3 folds for cross validation, split CFD into 71 training images and 46 testing images as the setting of [9,11,23], and split DRIVE into 20 training images and 20 testing images as the setting of [18,19].…”

Section: Experiments 31 Experimental Setupmentioning

confidence: 99%

“…The batch size is set as 32. We adopt the data augmentation strategy in [9], which includes random flipping, rotation, cropping, and various photometric augmentations.…”

Section: Experiments 31 Experimental Setupmentioning

confidence: 99%

“…We report two metrics for crack datasets: Optimal Image Scale (OIS) and Optimal Dataset Scale (ODS). The definitions of these metrics can be refered in [9]. Unlike the previous works [4,9], we consider 0 tolerance when counting true positive (TP) pixels, i.e.…”

Section: Crack Segmentation On Cfd and Aiglern Datasetsmentioning

confidence: 99%

“…To address the aforementioned task, several methods designed for crack segmentation/detection have been proposed [4,5,6,7,8,9,10,11]. Zhou et al [4] propose Deep-Crack which learns multi-scale deep features to capture line Input Image Initial Pred.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Crack Detection and Refinement Via Deep Reinforcement Learning

Park

Chen

2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

Detecting small cracks in concrete is difficult due to the complexity and thinness of cracking patterns, which requires the development of refined vision-based segmentation algorithms that can accurately characterize the details of crack defects. While existing methods are good at generally outlining cracks, due to inherent differences in shape distributions between common objects and cracks, their predictions often have disconnected segments and inaccuracy along boundaries. To this end, we develop a refinement framework using reinforcement learning (RL) that can better recognize details specific to cracks. Our method uses an RL agent to iteratively improve per-pixel crack predictions of a general segmentation model. We find that in addition to connecting gaps in predictions, the RL agent is also able to detect cracks that are missed in the original predictions. It does so by using the originally detected regions as crack priors to branch out from. Refining outputs of a commonly used per-pixel segmentation model, our method outperforms the current state-of-theart approaches for crack segmentation. Our experiments also demonstrate that our method generalizes well to a similar task of vessel segmentation.

show abstract

Section: Ground Truthmentioning

confidence: 99%

Section: Experiments 31 Experimental Setupmentioning

confidence: 99%

“…The batch size is set as 32. We adopt the data augmentation strategy in [9], which includes random flipping, rotation, cropping, and various photometric augmentations.…”

Section: Experiments 31 Experimental Setupmentioning

confidence: 99%