# Importance-Aware Semantic Segmentation in Self-Driving with Discrete Wasserstein Training

arXiv: https://arxiv.org/abs/2010.12440 · HTML: https://mlsanity.com/paper/2010.12440/

## Key facts

- Cited by 27 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 Oct 2020.

## Built on by (most cited first, 27 of 27)

- [Deep Unsupervised Domain Adaptation: A Review of Recent Advances and Perspectives](https://mlsanity.com/paper/2208.07422/) (2022)
- [Recent Advances in Optimal Transport for Machine Learning](https://mlsanity.com/paper/2306.16156/) (2023)
- [FreeMask: Synthetic Images with Dense Annotations Make Stronger Segmentation Models](https://mlsanity.com/paper/2310.15160/) (2023)
- [Meta Optimal Transport](https://mlsanity.com/paper/2206.05262/) (2022)
- [Dual-Reference Source-Free Active Domain Adaptation for Nasopharyngeal Carcinoma Tumor Segmentation across Multiple Hospitals](https://mlsanity.com/paper/2309.13401/) (2023)
- [Energy-constrained Self-training for Unsupervised Domain Adaptation](https://mlsanity.com/paper/2101.00316/) (2021)
- [Recursively Conditional Gaussian for Ordinal Unsupervised Domain Adaptation](https://mlsanity.com/paper/2107.13467/) (2021)
- [Identity-aware Facial Expression Recognition in Compressed Video](https://mlsanity.com/paper/2101.00317/) (2021)
- [Unsupervised Domain Adaptation for Segmentation with Black-box Source Model](https://mlsanity.com/paper/2208.07769/) (2022)
- [Exploiting Shape Cues for Weakly Supervised Semantic Segmentation](https://mlsanity.com/paper/2208.04286/) (2022)
- [Embedding Semantic Hierarchy in Discrete Optimal Transport for Risk Minimization](https://mlsanity.com/paper/2105.00101/) (2021)
- [Variational Inference for Quantifying Inter-observer Variability in Segmentation of Anatomical Structures](https://mlsanity.com/paper/2201.07106/) (2022)
- [PEARL: Preprocessing Enhanced Adversarial Robust Learning of Image Deraining for Semantic Segmentation](https://mlsanity.com/paper/2305.15709/) (2023)
- [Safety Metrics for Semantic Segmentation in Autonomous Driving](https://mlsanity.com/paper/2105.10142/) (2021)
- [Structure-aware Unsupervised Tagged-to-Cine MRI Synthesis with Self Disentanglement](https://mlsanity.com/paper/2202.12474/) (2022)
- [Robust Category-Level 3D Pose Estimation from Synthetic Data](https://mlsanity.com/paper/2305.16124/) (2023)
- [Hierarchical Image Peeling: A Flexible Scale-space Filtering Framework](https://mlsanity.com/paper/2104.01534/) (2021)
- [UDA4Inst: Unsupervised Domain Adaptation for Instance Segmentation](https://mlsanity.com/paper/2405.09682/) (2024)
- [Dual form Complementary Masking for Domain-Adaptive Image Segmentation](https://mlsanity.com/paper/2507.12008/) (2025)
- [DPGLA: Bridging the Gap between Synthetic and Real Data for Unsupervised Domain Adaptation in 3D LiDAR Semantic Segmentation](https://mlsanity.com/paper/2510.23525/) (2025)
- [Comparison of atlas-based and neural-network-based semantic segmentation for DENSE MRI images](https://mlsanity.com/paper/2109.14116/) (2021)
- [High Efficiency Pedestrian Crossing Prediction](https://mlsanity.com/paper/2204.01862/) (2022)
- [Integrated In-vehicle Monitoring System Using 3D Human Pose Estimation and Seat Belt Segmentation](https://mlsanity.com/paper/2204.07946/) (2022)
- [P-NOC: adversarial training of CAM generating networks for robust weakly supervised semantic segmentation priors](https://mlsanity.com/paper/2305.12522/) (2023)
- [Is Generative Modeling-based Stylization Necessary for Domain Adaptation in Regression Tasks?](https://mlsanity.com/paper/2306.01706/) (2023)
- [centroIDA: Cross-Domain Class Discrepancy Minimization Based on Accumulative Class-Centroids for Imbalanced Domain Adaptation](https://mlsanity.com/paper/2308.10619/) (2023)
- [Should Bias be Eliminated? A General Framework to Use Bias for OOD Generalization](https://mlsanity.com/paper/2507.17001/) (2025)

## Abstract (arXiv)

Semantic segmentation (SS) is an important perception manner for self-driving cars and robotics, which classifies each pixel into a pre-determined class. The widely-used cross entropy (CE) loss-based deep networks has achieved significant progress w.r.t. the mean Intersection-over Union (mIoU). However, the cross entropy loss can not take the different importance of each class in an self-driving system into account. For example, pedestrians in the image should be much more important than the surrounding buildings when make a decisions in the driving, so their segmentation results are expected to be as accurate as possible. In this paper, we propose to incorporate the importance-aware inter-class correlation in a Wasserstein training framework by configuring its ground distance matrix. The ground distance matrix can be pre-defined following a priori in a specific task, and the previous importance-ignored methods can be the particular cases. From an optimization perspective, we also extend our ground metric to a linear, convex or concave increasing function w.r.t. pre-defined ground distance. We evaluate our method on CamVid and Cityscapes datasets with different backbones (SegNet, ENet, FCN and Deeplab) in a plug and play fashion. In our extenssive experiments, Wasserstein loss demonstrates superior segmentation performance on the predefined critical classes for safe-driving.