REVIEW 5 cited by
Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
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, we find these two paradigms have different characteristics: (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 single/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 have been released at https://github.com/qizekun/ReCon.
Forward citations
Cited by 5 Pith papers
-
A Cross Branch Fusion-Based Contrastive Learning Framework for Point Cloud Self-supervised Learning
PoCCA improves point cloud self-supervised learning by fusing online and target branch features via cross-attention before the contrastive loss, achieving state-of-the-art among methods without extra training data.
-
Towards More Diverse and Challenging Pre-training for Point Cloud Learning: Self-Supervised Cross Reconstruction with Decoupled Views
Point-PQAE pre-trains point cloud transformers by cross-reconstructing one randomly cropped and rotated view from another, improving frozen-feature accuracy on ScanObjectNN by up to 7% over Point-MAE.
-
Asymmetric Dual Self-Distillation for 3D Self-Supervised Representation Learning
AsymDSD unifies latent masked point modeling and cross-view invariance self-distillation to learn 3D representations, reporting 90.53% on ScanObjectNN and 93.72% with 930k-shape pretraining.
-
UniPre3D: Unified Pre-training of 3D Point Cloud Models with Cross-Modal Gaussian Splatting
A point cloud pre-training method that uses 3D Gaussian splatting rendering and cross-modal image features to work for both objects and scenes.
-
PointLAMA: Latent Attention meets Mamba for Efficient Point Cloud Pretraining
PointLAMA inserts a point-wise latent attention block into a Mamba point cloud encoder and adds feature-space diffusion pretraining, reporting modest benchmark gains over PointMamba.
Discussion (0). Continue with ORCID to comment.