TY - GEN
T1 - Single image water hazard detection using FCN with reflection attention units
AU - Han, Xiaofeng
AU - Nguyen, Chuong
AU - You, Shaodi
AU - Lu, Jianfeng
N1 - Publisher Copyright:
© Springer Nature Switzerland AG 2018.
PY - 2018
Y1 - 2018
N2 - Water bodies, such as puddles and flooded areas, on and off road pose significant risks to autonomous cars. Detecting water from moving camera is a challenging task as water surface is highly refractive, and its appearance varies with viewing angle, surrounding scene, weather conditions. In this paper, we present a water puddle detection method based on a Fully Convolutional Network (FCN) with our newly proposed Reflection Attention Units (RAUs). An RAU is a deep network unit designed to embody the physics of reflection on water surface from sky and nearby scene. To verify the performance of our proposed method, we collect 11455 color stereo images with polarizers, and 985 of left images are annotated and divided into 2 datasets: On Road (ONR) dataset and Off Road (OFR) dataset. We show that FCN-8s with RAUs improves significantly precision and recall metrics as compared to FCN-8s, DeepLab V2 and Gaussian Mixture Model (GMM). We also show that focal loss function can improve the performance of FCN-8s network due to the extreme imbalance of water versus ground classification problem.
AB - Water bodies, such as puddles and flooded areas, on and off road pose significant risks to autonomous cars. Detecting water from moving camera is a challenging task as water surface is highly refractive, and its appearance varies with viewing angle, surrounding scene, weather conditions. In this paper, we present a water puddle detection method based on a Fully Convolutional Network (FCN) with our newly proposed Reflection Attention Units (RAUs). An RAU is a deep network unit designed to embody the physics of reflection on water surface from sky and nearby scene. To verify the performance of our proposed method, we collect 11455 color stereo images with polarizers, and 985 of left images are annotated and divided into 2 datasets: On Road (ONR) dataset and Off Road (OFR) dataset. We show that FCN-8s with RAUs improves significantly precision and recall metrics as compared to FCN-8s, DeepLab V2 and Gaussian Mixture Model (GMM). We also show that focal loss function can improve the performance of FCN-8s network due to the extreme imbalance of water versus ground classification problem.
KW - Deep learning
KW - Fully convolutional network
KW - Reflection attention unit
KW - Road hazard detection
KW - Water puddle detection
UR - http://www.scopus.com/inward/record.url?scp=85055137365&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-01231-1_7
DO - 10.1007/978-3-030-01231-1_7
M3 - Conference contribution
SN - 9783030012304
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 105
EP - 121
BT - Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
A2 - Hebert, Martial
A2 - Weiss, Yair
A2 - Ferrari, Vittorio
A2 - Sminchisescu, Cristian
PB - Springer Verlag
T2 - 15th European Conference on Computer Vision, ECCV 2018
Y2 - 8 September 2018 through 14 September 2018
ER -