# KDMOS:Knowledge Distillation for Motion Segmentation

arXiv: https://arxiv.org/abs/2506.14130 · HTML: https://mlsanity.com/paper/2506.14130/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Detection and dense prediction distillation, one of 12 directions in Knowledge distillation.
- Builds on 22 of its 28 references that are in the map.
- First posted on arXiv 17 Jun 2025.

## Placed in

- Knowledge distillation / [Detection and dense prediction distillation](https://mlsanity.com/distill/detection-and-dense-prediction-distillation/)

## Builds on (most cited first, 22 of 22)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Improved Knowledge Distillation via Teacher Assistant](https://mlsanity.com/paper/1902.03393/) (2019)
- [Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Segmentation](https://mlsanity.com/paper/2011.10033/) (2020)
- [Decoupled Knowledge Distillation](https://mlsanity.com/paper/2203.08679/) (2022)
- [On the Efficacy of Knowledge Distillation](https://mlsanity.com/paper/1910.01348/) (2019)
- [2DPASS: 2D Priors Assisted Semantic Segmentation on LiDAR Point Clouds](https://mlsanity.com/paper/2207.04397/) (2022)
- [Point-to-Voxel Knowledge Distillation for LiDAR Semantic Segmentation](https://mlsanity.com/paper/2206.02099/) (2022)
- [Logit Standardization in Knowledge Distillation](https://mlsanity.com/paper/2403.01427/) (2024)
- [One-for-All: Bridge the Gap Between Heterogeneous Architectures in Knowledge Distillation](https://mlsanity.com/paper/2310.19444/) (2023)
- [Moving Object Segmentation in 3D LiDAR Data: A Learning-based Approach Exploiting Sequential Data](https://mlsanity.com/paper/2105.08971/) (2021)
- [Learning to Upsample by Learning to Sample](https://mlsanity.com/paper/2308.15085/) (2023)
- [Receding Moving Object Segmentation in 3D LiDAR Data Using Sparse 4D Convolutions](https://mlsanity.com/paper/2206.04129/) (2022)
- [Efficient Spatial-Temporal Information Fusion for LiDAR-Based 3D Moving Object Segmentation](https://mlsanity.com/paper/2207.02201/) (2022)
- [Automatic Labeling to Generate Training Data for Online LiDAR-based Moving Object Segmentation](https://mlsanity.com/paper/2201.04501/) (2022)
- [MF-MOS: A Motion-Focused Model for Moving Object Segmentation](https://mlsanity.com/paper/2401.17023/) (2024)
- [InsMOS: Instance-Aware Moving Object Segmentation in LiDAR Data](https://mlsanity.com/paper/2303.03909/) (2023)
- [LiMoSeg: Real-time Bird's Eye View based LiDAR Motion Segmentation](https://mlsanity.com/paper/2111.04875/) (2021)
- [Self-Supervised Learning of Lidar Segmentation for Autonomous Indoor Navigation](https://mlsanity.com/paper/2012.05897/) (2020)
- [MambaMOS: LiDAR-based 3D Moving Object Segmentation with Motion-aware State Space Model](https://mlsanity.com/paper/2404.12794/) (2024)
- [MotionBEV: Attention-Aware Online LiDAR Moving Object Segmentation with Bird's Eye View based Appearance and Motion Features](https://mlsanity.com/paper/2305.07336/) (2023)
- [Knowledge Distillation from 3D to Bird's-Eye-View for LiDAR Semantic Segmentation](https://mlsanity.com/paper/2304.11393/) (2023)
- [CV-MOS: A Cross-View Model for Motion Segmentation](https://mlsanity.com/paper/2408.13790/) (2024)

## Abstract (arXiv)

Motion Object Segmentation (MOS) is crucial for autonomous driving, as it enhances localization, path planning, map construction, scene flow estimation, and future state prediction. While existing methods achieve strong performance, balancing accuracy and real-time inference remains a challenge. To address this, we propose a logits-based knowledge distillation framework for MOS, aiming to improve accuracy while maintaining real-time efficiency. Specifically, we adopt a Bird's Eye View (BEV) projection-based model as the student and a non-projection model as the teacher. To handle the severe imbalance between moving and non-moving classes, we decouple them and apply tailored distillation strategies, allowing the teacher model to better learn key motion-related features. This approach significantly reduces false positives and false negatives. Additionally, we introduce dynamic upsampling, optimize the network architecture, and achieve a 7.69% reduction in parameter count, mitigating overfitting. Our method achieves a notable IoU of 78.8% on the hidden test set of the SemanticKITTI-MOS dataset and delivers competitive results on the Apollo dataset. The KDMOS implementation is available at https://github.com/SCNU-RISLAB/KDMOS.