An Iteratively Optimized Patch Label Inference Network for Automatic Pavement Distress Detection

We present a novel deep learning framework named the Iteratively Optimized Patch Label Inference Network (IOPLIN) for automatically detecting various pavement distresses that are not solely limited to specific ones, such as cracks and potholes. IOPLIN can be iteratively trained with only image label via Expectation-Maximization Inspired Distillation (EMIPLD) strategy, accomplish this task well by inferring labels of patches from images. enjoys many desirable properties over state-of-the-art single branch CNN models GoogLeNet EfficientNet. It is able handle images in different resolutions, sufficiently utilize information particularly high-resolution since extracts visual features unrevised instead resized entire image. Moreover, it roughly localize distress without using any prior localization training phase. In order better evaluate effectiveness our method practice, we construct large-scale Bituminous Pavement Disease Detection dataset CQU-BPDD consisting 60,059 images, which acquired areas at times. Extensive results on demonstrate superiority classification approaches automatic detection. The source codes released https://github.com/DearCaat/ioplin , accessed xmlns:xlink="http://www.w3.org/1999/xlink">https://dearcaat.github.io/CQU-BPDD/ .