mlsanityData to 14 Oct 2026

A Realism Metric for Generated LiDAR Point Clouds

arXiv 2208.14958 ↗First posted 31 Aug 2022cs.CV cs.AI cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

A considerable amount of research is concerned with the generation of realistic sensor data. LiDAR point clouds are generated by complex simulations or learned generative models. The generated data is usually exploited to enable or improve downstream perception algorithms. Two major questions arise from these procedures: First, how to evaluate the realism of the generated data? Second, does more realistic data also lead to better perception performance? This paper addresses both questions and presents a novel metric to quantify the realism of LiDAR point clouds. Relevant features are learned from real-world and synthetic point clouds by training on a proxy classification task. In a series of experiments, we demonstrate the application of our metric to determine the realism of generated LiDAR data and compare the realism estimation of our metric to the performance of a segmentation model. We confirm that our metric provides an indication for the downstream segmentation performance.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 48 of its 64 references that are in the map. First posted on arXiv 31 Aug 2022.

Builds on

48 of 64 references.

References · 48

  1. 2022Generative Adversarial Networkscited by 5,627
  2. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,248
  3. 2016Image-to-Image Translation with Conditional Adversarial Networkscited by 4,527
  4. 2019nuScenes: A multimodal dataset for autonomous drivingcited by 4,523
  5. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,401
  6. 2016Improved Techniques for Training GANscited by 2,970
  7. 2015ShapeNet: An Information-Rich 3D Model Repositorycited by 2,897
  8. 2016Perceptual Losses for Real-Time Style Transfer and Super-Resolutioncited by 2,694
  9. 2017CARLA: An Open Urban Driving Simulatorcited by 2,449
  10. 2015Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networkscited by 2,054
  11. 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,908
  12. 2019Semantic Image Synthesis with Spatially-Adaptive Normalizationcited by 1,299
36 more
  1. 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 860
  2. 2017Wasserstein GANcited by 639
  3. 2016StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networkscited by 577
  4. 2014Deep metric learning using Triplet networkcited by 422
  5. 2015A note on the evaluation of generative modelscited by 292
  6. 2017Are GANs Created Equal? A Large-Scale Studycited by 267
  7. 2018SqueezeSegV2: Improved Model Structure and Unsupervised Domain Adaptation for Road-Object Segmentation from a LiDAR Point Cloudcited by 228
  8. 2018Pros and Cons of GAN Evaluation Measurescited by 224
  9. 2017VEEGAN: Reducing Mode Collapse in GANs using Implicit Variational Learningcited by 200
  10. 20193D Point Cloud Generative Adversarial Network Based on Tree Structured Graph Convolutionscited by 173
  11. 2017Data Decisions and Theoretical Implications when Adversarially Learning Fair Representationscited by 165
  12. 2016Mode Regularized Generative Adversarial Networkscited by 123
  13. 2017PacGAN: The power of two samples in generative adversarial networkscited by 104
  14. 2022BigDatasetGAN: Synthesizing ImageNet with Pixel-wise Annotationscited by 99
  15. 2016Disentangling factors of variation in deep representations using adversarial trainingcited by 87
  16. 2018An empirical study on evaluation metrics of generative adversarial networkscited by 81
  17. 2017DeLiGAN : Generative Adversarial Networks for Diverse and Limited Datacited by 77
  18. 2018Deep Generative Modeling of LiDAR Datacited by 71
  19. 2018On GANs and GMMscited by 62
  20. 2018Geometry Score: A Method For Comparing Generative Adversarial Networkscited by 57
  21. 2021A Survey on Deep Domain Adaptation for LiDAR Perceptioncited by 47
  22. 2016Stacked Generative Adversarial Networkscited by 44
  23. 2017LR-GAN: Layered Recursive Generative Adversarial Networks for Image Generationcited by 36
  24. 2017AdaGAN: Boosting Generative Modelscited by 34
  25. 2019Precise Synthetic Image and LiDAR (PreSIL) Dataset for Autonomous Vehicle Perceptioncited by 29
  26. 2019LiDAR Sensor modeling and Data augmentation with GANs for Autonomous drivingcited by 22
  27. 2018Gradient Reversal Against Discriminationcited by 20
  28. 2017On the Effects of Batch and Weight Normalization in Generative Adversarial Networkscited by 13
  29. 2016Generating images with recurrent adversarial networkscited by 9
  30. 2016Ensembles of Generative Adversarial Networkscited by 8
  31. 2019CNN-based synthesis of realistic high-resolution LiDAR datacited by 7
  32. 2021Self-Ensembling GAN for Cross-Domain Semantic Segmentationcited by 7
  33. 2018Decoupled Learning for Conditional Adversarial Networkscited by 5
  34. 2018Skill Rating for Generative Modelscited by 4
  35. 2021Quantifying point cloud realism through adversarially learned latent representationscited by 2
  36. 2022Improving Performance of Semantic Segmentation CycleGANs by Noise Injection into the Latent Segmentation Spacecited by 2

Built on by

Cited by year

Citing papers per year: 2023: 2; 2024: 0; 2025: 3.

5 papers.

  1. 2023Quantifying the LiDAR Sim-to-Real Domain Shift: A Detailed Investigation Using Object Detectors and Analyzing Point Clouds at Target-Levelcited by 7
  2. 2025Synth It Like KITTI: Synthetic Data Generation for Object Detection in Driving Scenarioscited by 3
  3. 2025Rethinking Metrics and Diffusion Architecture for 3D Point Cloud Generationcited by 2
  4. 2023Differentiable SLAM Helps Deep Learning-based LiDAR Perception Taskscited by 1
  5. 2025Mitigating Context Bias in Domain Adaptation for Object Detection using Mask Poolingcited by 0