Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Self-supervised Monocular Depth Estimation with Large Kernel Attention

Xuezhi Xiang, Yao Wang, Lei Zhang, Denis Ombati, Himaloy Himu, Xiantong Zhen

TL;DR

This paper proposes a decoder based on large kernel attention, which can model long-distance dependencies without compromising the two-dimension structure of features while maintaining feature channel adaptivity, and introduces a up-sampling module to accurately recover the fine details in the depth map.

Abstract

Self-supervised monocular depth estimation has emerged as a promising approach since it does not rely on labeled training data. Most methods combine convolution and Transformer to model long-distance dependencies to estimate depth accurately. However, Transformer treats 2D image features as 1D sequences, and positional encoding somewhat mitigates the loss of spatial information between different feature blocks, tending to overlook channel features, which limit the performance of depth estimation. In this paper, we propose a self-supervised monocular depth estimation network to get finer details. Specifically, we propose a decoder based on large kernel attention, which can model long-distance dependencies without compromising the two-dimension structure of features while maintaining feature channel adaptivity. In addition, we introduce a up-sampling module to accurately recover the fine details in the depth map. Our method achieves competitive results on the KITTI dataset.

Self-supervised Monocular Depth Estimation with Large Kernel Attention

TL;DR

This paper proposes a decoder based on large kernel attention, which can model long-distance dependencies without compromising the two-dimension structure of features while maintaining feature channel adaptivity, and introduces a up-sampling module to accurately recover the fine details in the depth map.

Abstract

Self-supervised monocular depth estimation has emerged as a promising approach since it does not rely on labeled training data. Most methods combine convolution and Transformer to model long-distance dependencies to estimate depth accurately. However, Transformer treats 2D image features as 1D sequences, and positional encoding somewhat mitigates the loss of spatial information between different feature blocks, tending to overlook channel features, which limit the performance of depth estimation. In this paper, we propose a self-supervised monocular depth estimation network to get finer details. Specifically, we propose a decoder based on large kernel attention, which can model long-distance dependencies without compromising the two-dimension structure of features while maintaining feature channel adaptivity. In addition, we introduce a up-sampling module to accurately recover the fine details in the depth map. Our method achieves competitive results on the KITTI dataset.
Paper Structure (11 sections, 4 equations, 5 figures, 2 tables)

This paper contains 11 sections, 4 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Overall architecture
  4. Large kernel attention
  5. Upsample module
  6. Experiments
  7. Datasets and Metrics
  8. Implementation details
  9. Results
  10. Ablation study
  11. Conclusion

Figures (5)

  • Figure 1: The overall architecture of our self-supervised monocular depth estimation method, which contains a depth network and a pose network.
  • Figure 2: Overview of our depth decoder. The features are fed into ${3\times3}$ convolution layer and are concatenated with the upsampled features of the next layer and fed into LKA.
  • Figure 3: The architecture of large kernel attention (LKA). It is composed with depthwise convolution, depth-wise dilation convolution, and ${1\times1}$ convolution.
  • Figure 4: The architecture of upsample module. The grid sample function uses the offset to resample ${F}^{'}_{in}$ to ${F}^{'}_{out}$.
  • Figure 5: Qualitative results on the KITTI dataset.Our model can obtain higher quality depth maps with finer depth edges compared to other methods.