# Unsupervised Online 3D Instance Segmentation with Synthetic Sequences and Dynamic Loss

arXiv: https://arxiv.org/abs/2509.23194 · HTML: https://mlsanity.com/paper/2509.23194/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 33 of its 46 references that are in the map.
- First posted on arXiv 27 Sep 2025.

## Builds on (most cited first, 30 of 33)

- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [nuScenes: A multimodal dataset for autonomous driving](https://mlsanity.com/paper/1903.11027/) (2019)
- [CARLA: An Open Urban Driving Simulator](https://mlsanity.com/paper/1711.03938/) (2017)
- [4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks](https://mlsanity.com/paper/1904.08755/) (2019)
- [PointGroup: Dual-Set Point Grouping for 3D Instance Segmentation](https://mlsanity.com/paper/2004.01658/) (2020)
- [Mask3D: Mask Transformer for 3D Semantic Instance Segmentation](https://mlsanity.com/paper/2210.03105/) (2022)
- [3D-SIS: 3D Semantic Instance Segmentation of RGB-D Scans](https://mlsanity.com/paper/1812.07003/) (2018)
- [PandaSet: Advanced Sensor Suite Dataset for Autonomous Driving](https://mlsanity.com/paper/2112.12610/) (2021)
- [SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation](https://mlsanity.com/paper/1711.08588/) (2017)
- [LiDARsim: Realistic LiDAR Simulation by Leveraging the Real World](https://mlsanity.com/paper/2006.09348/) (2020)
- [Segment Any Point Cloud Sequences by Distilling Vision Foundation Models](https://mlsanity.com/paper/2306.09347/) (2023)
- [LaserMix for Semi-Supervised LiDAR Semantic Segmentation](https://mlsanity.com/paper/2207.00026/) (2022)
- [Learning to Generate Realistic LiDAR Point Clouds](https://mlsanity.com/paper/2209.03954/) (2022)
- [DyCo3D: Robust Instance Segmentation of 3D Point Clouds through Dynamic Convolution](https://mlsanity.com/paper/2011.13328/) (2020)
- [PolarMix: A General Data Augmentation Technique for LiDAR Point Clouds](https://mlsanity.com/paper/2208.00223/) (2022)
- [4D Panoptic LiDAR Segmentation](https://mlsanity.com/paper/2102.12472/) (2021)
- [Patchwork++: Fast and Robust Ground Segmentation Solving Partial Under-Segmentation Using 3D Point Cloud](https://mlsanity.com/paper/2207.11919/) (2022)
- [VideoCutLER: Surprisingly Simple Unsupervised Video Instance Segmentation](https://mlsanity.com/paper/2308.14710/) (2023)
- [UnScene3D: Unsupervised 3D Instance Segmentation for Indoor Scenes](https://mlsanity.com/paper/2303.14541/) (2023)
- [Patchwork: Concentric Zone-based Region-wise Ground Segmentation with Ground Likelihood Estimation Using a 3D LiDAR Sensor](https://mlsanity.com/paper/2108.05560/) (2021)
- [GLENet: Boosting 3D Object Detectors with Generative Label Uncertainty Estimation](https://mlsanity.com/paper/2207.02466/) (2022)
- [Towards Unsupervised Object Detection From LiDAR Point Clouds](https://mlsanity.com/paper/2311.02007/) (2023)
- [LISO: Lidar-only Self-Supervised 3D Object Detection](https://mlsanity.com/paper/2403.07071/) (2024)
- [Rethinking Few-shot 3D Point Cloud Semantic Segmentation](https://mlsanity.com/paper/2403.00592/) (2024)
- [Generalized Few-shot 3D Point Cloud Segmentation with Vision-Language Model](https://mlsanity.com/paper/2503.16282/) (2025)
- [SeMoLi: What Moves Together Belongs Together](https://mlsanity.com/paper/2402.19463/) (2024)
- [PointMCD: Boosting Deep Point Cloud Encoders via Multi-view Cross-modal Distillation for 3D Shape Recognition](https://mlsanity.com/paper/2207.03128/) (2022)
- [Leveraging Single-View Images for Unsupervised 3D Point Cloud Completion](https://mlsanity.com/paper/2212.00564/) (2022)
- [Spatial-Temporal Graph Enhanced DETR Towards Multi-Frame 3D Object Detection](https://mlsanity.com/paper/2307.00347/) (2023)
- [Learning from Mistakes: Self-Regularizing Hierarchical Representations in Point Cloud Semantic Segmentation](https://mlsanity.com/paper/2301.11145/) (2023)

## Abstract (arXiv)

Unsupervised online 3D instance segmentation is a fundamental yet challenging task, as it requires maintaining consistent object identities across LiDAR scans without relying on annotated training data. Existing methods, such as UNIT, have made progress in this direction but remain constrained by limited training diversity, rigid temporal sampling, and heavy dependence on noisy pseudo-labels. We propose a new framework that enriches the training distribution through synthetic point cloud sequence generation, enabling greater diversity without relying on manual labels or simulation engines. To better capture temporal dynamics, our method incorporates a flexible sampling strategy that leverages both adjacent and non-adjacent frames, allowing the model to learn from long-range dependencies as well as short-term variations. In addition, a dynamic-weighting loss emphasizes confident and informative samples, guiding the network toward more robust representations. Through extensive experiments on SemanticKITTI, nuScenes, and PandaSet, our method consistently outperforms UNIT and other unsupervised baselines, achieving higher segmentation accuracy and more robust temporal associations. The code will be publicly available at github.com/Eaphan/SFT3D.