Single Image Water Hazard Detection using FCN with Reflection Attention Units
Xiaofeng Han ¹, Chuong Nguyen ²³⁴, Shaodi You ²³, Jianfeng Lu ^1
1School of Computer Science and Engineering, Nanjing University of Science and Technology
²Australian National University
³CSIRO DATA61, Australia
⁴Australian Centre of Excellence for Robotic Vision
ECCV 2018

This is a work in progress and the final codes are coming soon.

The figure below shows one example of water hazard detection. In this figure, the red areas are the water areas.

The figure shows our network architecture. The figure below shows our proposed Reflection Attention Units

The paper can be downloaded here: Single Image Water Hazard Detection using FCN with Reflection Attention Units

The demo videos can be downloaded here:
ECCV DEMO video
OFF ROAD
ON ROAD

The dataset used in our paper can be downloaded here:
Puddle-1000

The codes are modified based on this project: https://github.com/shekkizh/FCN.tensorflow, and we'd like to thank the authors very much.