DSpace Repository

Semantic Segmentation of Bridge Inspection Images for Damage Assessment

Show simple item record

dc.contributor.advisor Dailey, Matthew
dc.contributor.author Laiteerapong, Teera
dc.contributor.other Ekpanyapong, Mongkol
dc.contributor.other Ouivirach, Kan
dc.date.accessioned 2019-05-07T04:29:59Z
dc.date.available 2019-05-07T04:29:59Z
dc.date.issued 2019-05
dc.identifier.other AIT
dc.identifier.uri http://www.cs.ait.ac.th/xmlui/handle/123456789/936
dc.description 79 p. en_US
dc.description.abstract Bridge maintenance required monitoring. Otherwise, a severe accident could happen. In general, the bridge maintenance required an inspector to go to the bridge, take a photo, and collect them at the government office. The procedures require an inspector to determine the type of damage corresponding to the photos. This research work attempts to assist the inspector by automatically labeling the damage type on the particular area in the image. The approaches begin with fully convolutional network as a baseline mode, the state-of-the-art model Mask R-CNN, and hyperparameters fine-tuning. The dataset contains two classes of the bridge damage on the deck area. The first one is delamination, and the second one is rebar exposure. The exist of rebar exposure can be interpret as a severe case. Therefore, false negative is as crucial as false positive in this kind of dataset. This dataset contains three experts labelled. However, there is much disagreement among their labels. The mixture of experts using non-maximum suppression helps obtain more recall. Lastly, with all the damage label information, the government can plan the bridge maintenance more effectively. en_US
dc.description.sponsorship Royal Thai Government - AIT Fellowship en_US
dc.publisher AIT en_US
dc.subject Damage Detection en_US
dc.subject Delamination en_US
dc.subject Image Segmentation en_US
dc.subject Mixture of Experts en_US
dc.subject Rebar Exposure en_US
dc.subject Semantics en_US
dc.title Semantic Segmentation of Bridge Inspection Images for Damage Assessment en_US
dc.type Thesis en_US

Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace

Advanced Search


My Account