Efficient deep models for monocular road segmentation

This paper addresses the problem of road scene segmentation in conventional RGB images by exploiting recent advances in semantic segmentation via convolutional neural networks (CNNs). Segmentation networks are very large and do not currently run at interactive frame rates. To make this technique applicable to robotics we propose several architecture refinements that provide the best trade-off between segmentation quality and runtime. This is achieved by a new mapping between classes and filters at the expansion side of the network. The network is trained end-to-end and yields precise road/lane predictions at the original input resolution in roughly 50ms. Compared to the state of the art, the network achieves top accuracies on the KITTI dataset for road and lane segmentation while providing a 20× speed-up. We demonstrate that the improved efficiency is not due to the road segmentation task. Also on segmentation datasets with larger scene complexity, the accuracy does not suffer from the large speed-up.

Authors

• Gabriel L. Oliveira
• Wolfram Burgard
• Thomas Brox

Keywords

• Roads
• Image segmentation
• Computer architecture
• Training
• Semantics
• Robots
• Runtime
• Road Segments
• Convolutional Network
• Convolutional Neural Network
• Semantic Segmentation
• Segmentation Task
• KITTI Dataset
• Street Scenes
• Network Side
• Deep Learning
• Deep Network
• Convolutional Layers
• Part Of Network
• Data Augmentation
• Stochastic Gradient Descent
• Training Images
• Classification Network
• False Negative Rate
• Network Output
• Reduction In Parameters
• Number Of Filters
• Conditional Random Field
• Fully Convolutional Network
• Contractive
• Segmentation Problem
• Convolutional Neural Network Classifier
• Impact Of Reduction
• Forward Pass
• Fixed Learning Rate
• Distribution Of Parameters
• False Positive Rate