본문 바로가기
카테고리 없음

[자율주행 차선관련 논문] 2022, CLRNet: Cross Layer Refinement Network for Lane Detection

by icebear3000 2023. 3. 30.
반응형

Abstract

 

1. Introduction

 

2. Related Work

2.1. Segmentation-based methods

 

2.2. Anchor-based methods

 

2.3. Parameter-based methods

 

3. Approach

3.1. The Lane Representation

Lane Prior.

 

3.2. Cross Layer Refinement

Motivation

 

Refinement structure.

 

3.3. ROIGather

Motivation

 

ROIGather structure

 

 

3.4. Line IoU loss

Motivation

 

Formula

 

 

3.5. Training and Infercence Details

Positive samples selection.

 

Training Loss.

 

 

4. Experiment
4.1. Datasets

 

4.2. Implementation details

 

 

4.3. Evaluation Metric

 

 

4.4. Comparison with the state-of-the-art results

Performance on LLAMAS.

 

Performance on Tusimple.

 

4.5. Ablation study

 

Overall Ablation Study. 

 

Analysis for ROIGather.

 

Ablation study on Cross Layer Refinement.

 

Ablation Study on Line-IoU Loss. 

 

 

5. Conclusion

반응형

댓글