The following metrics are consistently used in our benchmark:
-
Mean Corruption Error (mCE):
- The Corruption Error (CE) for model
$A$ under corruption type$i$ across 3 severity levels is:$\text{CE}_i^{\text{Model}A} = \frac{\sum((1 - \text{mIoU})^{\text{Model}A})}{\sum((1 - \text{mIoU})^{\text{Baseline}})}$ . - The average CE for model
$A$ on all$N$ corruption types, i.e., mCE, is calculated as:$\text{mCE} = \frac{1}{N}\sum\text{CE}_i$ .
- The Corruption Error (CE) for model
-
Mean Resilience Rate (mRR):
- The Resilience Rate (RR) for model
$A$ under corruption type$i$ across 3 severity levels is:$\text{RR}_i^{\text{Model}A} = \frac{\sum(\text{mIoU}^{\text{Model}A})}{3\times (\text{clean-mIoU}^{\text{Model}A})} .$ - The average RR for model
$A$ on all$N$ corruption types, i.e., mRR, is calculated as:$\text{mRR} = \frac{1}{N}\sum\text{RR}_i$ .
- The Resilience Rate (RR) for model
Corruption | Light | Moderate | Heavy | Average | ||
---|---|---|---|---|---|---|
Fog | 45.80 | 44.84 | 37.47 | 42.70 | 129.84 | 68.27 |
Wet Ground | 60.67 | 56.35 | 54.99 | 57.34 | 92.72 | 91.67 |
Snow | 55.53 | 53.85 | 51.55 | 53.64 | 99.23 | 85.76 |
Motion Blur | 56.92 | 52.87 | 48.35 | 52.71 | 70.50 | 84.27 |
Beam Missing | 61.40 | 56.67 | 49.28 | 55.78 | 101.24 | 89.18 |
Crosstalk | 48.81 | 45.43 | 41.87 | 45.37 | 131.13 | 72.53 |
Incomplete Echo | 57.77 | 54.25 | 48.19 | 53.40 | 102.26 | 85.37 |
Cross-Sensor | 58.16 | 51.34 | 28.01 | 45.84 | 100.39 | 73.29 |
-
Summary:
$\text{mIoU}_{\text{clean}} =$ 62.55%,$\text{mCE} =$ 103.41%,$\text{mRR} =$ 81.29%.
Corruption | Light | Moderate | Heavy | Average | ||
---|---|---|---|---|---|---|
Fog | 68.49 | 67.98 | 64.56 | 67.01 | ||
Wet Ground | 71.51 | 70.23 | 67.86 | 69.87 | ||
Snow | ||||||
Motion Blur | 63.99 | 58.75 | 52.18 | 58.31 | ||
Beam Missing | 58.57 | 49.11 | 42.23 | 49.97 | ||
Crosstalk | ||||||
Incomplete Echo | 56.53 | 52.99 | 50.40 | 53.31 | ||
Cross-Sensor |
-
Summary:
$\text{mIoU}_{\text{clean}} =$ 73.28%,$\text{mCE} =$ %,$\text{mRR} =$ %.
@inproceedings{cheng2022cenet,
title = {CENet: Toward Concise and Efficient Lidar Semantic Segmentation for Autonomous Driving},
author = {Cheng, Hui-Xian and Han, Xian-Feng and Xiao, Guo-Qiang},
booktitle = {IEEE International Conference on Multimedia and Expo},
year = {2022},
}