Masked Discrimination for Self-supervised Learning on Point Clouds

Liu, Haotian; Cai, Mu; Lee, Yong Jae

doi:10.1007/978-3-031-20086-1_38

Cited by 51 publications

(35 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…PointM2AE (Zhang et al, 2022b) uses a hierarchical Transformer and designs the corresponding masking strategy. MaskPoint (Liu et al, 2022) proposes to add some noise points and classify whether they belong to masking tokens. Recently, ACT (Dong et al, 2023) uses a cross-modal autoencoder as the reconstruction target to acquire dark knowledge from other modalities.…”

Section: Related Workmentioning

confidence: 99%

Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining

Qi¹,

Dong²,

Fan³

et al. 2023

Preprint

View full text Add to dashboard Cite

Mainstream 3D representation learning approaches are built upon contrastive or generative modeling pretext tasks, where great improvements in performance on various downstream tasks have been achieved. However, by investigating the methods of these two paradigms, we find that (i) contrastive models are data-hungry that suffer from a representation over-fitting issue; (ii) generative models have a data filling issue that shows inferior data scaling capacity compared to contrastive models. This motivates us to learn 3D representations by sharing the merits of both paradigms, which is non-trivial due to the pattern difference between the two paradigms. In this paper, we propose contrast with reconstruct (RECON) that unifies these two paradigms. RECON is trained to learn from both generative modeling teachers and cross-modal contrastive teachers through ensemble distillation, where the generative student guides the contrastive student. An encoder-decoder style RECON-block is proposed that transfers knowledge through cross attention with stop-gradient, which avoids pretraining over-fitting and pattern difference issues. RECON achieves a new state-of-the-art in 3D representation learning, e.g., 91.26% accuracy on ScanObjectNN. Codes will be released at https://github.com/qizekun/ReCon.

show abstract

Section: Related Workmentioning

confidence: 99%

Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining

Qi¹,

Dong²,

Fan³

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The recent rise to prominence of ViT architectures in computer vision has inspired a series of works for indoor point cloud segmentation [29,54,60,64,65].…”

Section: Vits For Point Cloud Segmentationmentioning

confidence: 99%

RangeViT: Towards Vision Transformers for 3D Semantic Segmentation in Autonomous Driving

Ando¹,

Gidaris²,

Bursuc³

et al. 2023

Preprint

View full text Add to dashboard Cite

Casting semantic segmentation of outdoor LiDAR point clouds as a 2D problem, e.g., via range projection, is an effective and popular approach. These projection-based methods usually benefit from fast computations and, when combined with techniques which use other point cloud representations, achieve state-of-the-art results. Today, projection-based methods leverage 2D CNNs but recent advances in computer vision show that vision transformers (ViTs) have achieved state-of-the-art results in many imagebased benchmarks. In this work, we question if projectionbased methods for 3D semantic segmentation can benefit from these latest improvements on ViTs. We answer positively but only after combining them with three key ingredients: (a) ViTs are notoriously hard to train and require a lot of training data to learn powerful representations. By preserving the same backbone architecture as for RGB images, we can exploit the knowledge from long training on large image collections that are much cheaper to acquire and annotate than point clouds. We reach our best results with pre-trained ViTs on large image datasets. (b) We compensate ViTs' lack of inductive bias by substituting a tailored convolutional stem for the classical linear embedding layer. (c) We refine pixel-wise predictions with a convolutional decoder and a skip connection from the convolutional stem to combine low-level but fine-grained features of the the convolutional stem with the high-level but coarse predictions of the ViT encoder. With these ingredients, we show that our method, called RangeViT, outperforms existing projection-based methods on nuScenes and SemanticKITTI. We provide the implementation code at https://github.com/valeoai/rangevit.

show abstract

“…Annotating point clouds demands significant effort, necessitating self-supervised pre-training methods. Prior approaches primarily focus on object CAD models [21,26,29,39,42,44] and indoor scenes [17,35,46]. Point-BERT [42] applies BERT-like paradigms for point cloud recognition, while Point-MAE [26] reconstructs point patches without the tokenizer.…”

Section: Related Workmentioning

confidence: 99%

MV-JAR: Masked Voxel Jigsaw and Reconstruction for LiDAR-Based Self-Supervised Pre-Training

Xu¹,

Wang²,

Zhang³

et al. 2023

Preprint

View full text Add to dashboard Cite

This paper introduces the Masked Voxel Jigsaw and Reconstruction (MV-JAR) method for LiDAR-based selfsupervised pre-training and a carefully designed dataefficient 3D object detection benchmark on the Waymo dataset. Inspired by the scene-voxel-point hierarchy in downstream 3D object detectors, we design masking and reconstruction strategies accounting for voxel distributions in the scene and local point distributions within the voxel. We employ a Reversed-Furthest-Voxel-Sampling strategy to address the uneven distribution of LiDAR points and propose MV-JAR, which combines two techniques for modeling the aforementioned distributions, resulting in superior performance. Our experiments reveal limitations in previous dataefficient experiments, which uniformly sample fine-tuning splits with varying data proportions from each LiDAR sequence, leading to similar data diversity across splits. To address this, we propose a new benchmark that samples scene sequences for diverse fine-tuning splits, ensuring adequate model convergence and providing a more accurate evaluation of pre-training methods. Experiments on our Waymo benchmark and the KITTI dataset demonstrate that MV-JAR consistently and significantly improves 3D detection performance across various data scales, achieving up to a 6.3% increase in mAPH compared to training from scratch. Codes and the benchmark will be available at https://github.com/SmartBot-PJLab/MV-JAR.

show abstract

Masked Discrimination for Self-supervised Learning on Point Clouds

Cited by 51 publications

References 43 publications

Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining

Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining

RangeViT: Towards Vision Transformers for 3D Semantic Segmentation in Autonomous Driving

MV-JAR: Masked Voxel Jigsaw and Reconstruction for LiDAR-Based Self-Supervised Pre-Training

Contact Info

Product

Resources

About