# Multi-Frame to Single-Frame: Knowledge Distillation for 3D Object Detection

arXiv: https://arxiv.org/abs/2009.11859 · HTML: https://mlsanity.com/paper/2009.11859/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 24 Sep 2020.

## Built on by (most cited first, 12 of 12)

- [Learning Distilled Collaboration Graph for Multi-Agent Perception](https://mlsanity.com/paper/2111.00643/) (2021)
- [Object DGCNN: 3D Object Detection using Dynamic Graphs](https://mlsanity.com/paper/2110.06923/) (2021)
- [PointDistiller: Structured Knowledge Distillation Towards Efficient and Compact 3D Detection](https://mlsanity.com/paper/2205.11098/) (2022)
- [Pseudo-labeling for Scalable 3D Object Detection](https://mlsanity.com/paper/2103.02093/) (2021)
- [Towards Efficient 3D Object Detection with Knowledge Distillation](https://mlsanity.com/paper/2205.15156/) (2022)
- [Sparse2Dense: Learning to Densify 3D Features for 3D Object Detection](https://mlsanity.com/paper/2211.13067/) (2022)
- [Paint and Distill: Boosting 3D Object Detection with Semantic Passing Network](https://mlsanity.com/paper/2207.05497/) (2022)
- [Boosting Single-Frame 3D Object Detection by Simulating Multi-Frame Point Clouds](https://mlsanity.com/paper/2207.01030/) (2022)
- [Label-Guided Auxiliary Training Improves 3D Object Detector](https://mlsanity.com/paper/2207.11753/) (2022)
- [V2X-AHD:Vehicle-to-Everything Cooperation Perception via Asymmetric Heterogenous Distillation Network](https://mlsanity.com/paper/2310.06603/) (2023)
- [Semantic-Supervised Spatial-Temporal Fusion for LiDAR-based 3D Object Detection](https://mlsanity.com/paper/2503.10579/) (2025)
- [Longitudinal Risk Prediction in Mammography with Privileged History Distillation](https://mlsanity.com/paper/2603.15814/) (2026)

## Abstract (arXiv)

A common dilemma in 3D object detection for autonomous driving is that high-quality, dense point clouds are only available during training, but not testing. We use knowledge distillation to bridge the gap between a model trained on high-quality inputs at training time and another tested on low-quality inputs at inference time. In particular, we design a two-stage training pipeline for point cloud object detection. First, we train an object detection model on dense point clouds, which are generated from multiple frames using extra information only available at training time. Then, we train the model's identical counterpart on sparse single-frame point clouds with consistency regularization on features from both models. We show that this procedure improves performance on low-quality data during testing, without additional overhead.