Improving Monocular Depth Estimation by Semantic Pre-training

Knowing the distance to nearby objects is crucial for autonomous cars to navigate safely in everyday traffic. In this paper, we investigate monocular depth estimation, which advanced substantially within the last years and is providing increasingly more accurate results while only requiring a single camera image as input. In line with recent work, we use an encoder-decoder structure with so-called packing layers to estimate depth values in a self-supervised fashion. We propose integrating a joint pre-training of semantic segmentation plus depth estimation on a dataset providing semantic labels. By using a separate semantic decoder that is only needed for pre-training, we can keep the network comparatively small. Our extensive experimental evaluation shows that the addition of such pre-training improves the depth estimation performance substantially. Finally, we show that we achieve competitive performance on the KITTI dataset despite using a much smaller and more efficient network.

Authors

• Peter Rottmann
• Thorbjörn Posewsky
• Andres Milioto
• Cyrill Stachniss
• Jens Behley

Keywords

• Image segmentation
• Navigation
• Semantics
• Buildings
• Estimation
• Computer architecture
• Cameras
• Depth Estimation
• Monocular Depth Estimation
• Single Image
• Semantic Segmentation
• Depth Values
• Semantic Labels
• Encoder-decoder Structure
• KITTI Dataset
• Semantic Information
• Image Pairs
• Target Image
• Depth Map
• Source Images
• Ground Truth Labels
• Inference Time
• Pre-trained Network
• Semantic Network
• Self-supervised Learning
• Stereopsis
• Depth Perception
• Stereo Images
• Stereo Pairs
• Pretext Task
• Matching Loss
• Relative Absolute Error
• Depth Prediction
• Monocular Images
• Warped Image
• Multiple Tasks
• Target Dataset