# ConDense: Consistent 2D/3D Pre-training for Dense and Sparse Features from Multi-View Images

arXiv: https://arxiv.org/abs/2408.17027 · HTML: https://mlsanity.com/paper/2408.17027/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 66 of its 78 references that are in the map.
- First posted on arXiv 30 Aug 2024.

## Builds on (most cited first, 30 of 66)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Momentum Contrast for Unsupervised Visual Representation Learning](https://mlsanity.com/paper/1911.05722/) (2019)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Emerging Properties in Self-Supervised Vision Transformers](https://mlsanity.com/paper/2104.14294/) (2021)
- [PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation](https://mlsanity.com/paper/1612.00593/) (2016)
- [PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space](https://mlsanity.com/paper/1706.02413/) (2017)
- [ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes](https://mlsanity.com/paper/1702.04405/) (2017)
- [ShapeNet: An Information-Rich 3D Model Repository](https://mlsanity.com/paper/1512.03012/) (2015)
- [Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fields](https://mlsanity.com/paper/2111.12077/) (2021)
- [pixelNeRF: Neural Radiance Fields from One or Few Images](https://mlsanity.com/paper/2012.02190/) (2020)
- [3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation](https://mlsanity.com/paper/1606.06650/) (2016)
- [Context Encoders: Feature Learning by Inpainting](https://mlsanity.com/paper/1604.07379/) (2016)
- [SuperPoint: Self-Supervised Interest Point Detection and Description](https://mlsanity.com/paper/1712.07629/) (2017)
- [4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks](https://mlsanity.com/paper/1904.08755/) (2019)
- [KPConv: Flexible and Deformable Convolution for Point Clouds](https://mlsanity.com/paper/1904.08889/) (2019)
- [Colorful Image Colorization](https://mlsanity.com/paper/1603.08511/) (2016)
- [The Replica Dataset: A Digital Replica of Indoor Spaces](https://mlsanity.com/paper/1906.05797/) (2019)
- [iBOT: Image BERT Pre-Training with Online Tokenizer](https://mlsanity.com/paper/2111.07832/) (2021)
- [MVSNet: Depth Inference for Unstructured Multi-view Stereo](https://mlsanity.com/paper/1804.02505/) (2018)
- [PointCNN: Convolution On X-Transformed Points](https://mlsanity.com/paper/1801.07791/) (2018)
- [MVSNeRF: Fast Generalizable Radiance Field Reconstruction from Multi-View Stereo](https://mlsanity.com/paper/2103.15595/) (2021)
- [Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modeling](https://mlsanity.com/paper/2111.14819/) (2021)
- [ImageNet-21K Pretraining for the Masses](https://mlsanity.com/paper/2104.10972/) (2021)
- [OpenScene: 3D Scene Understanding with Open Vocabularies](https://mlsanity.com/paper/2211.15654/) (2022)
- [LERF: Language Embedded Radiance Fields](https://mlsanity.com/paper/2303.09553/) (2023)
- [Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data](https://mlsanity.com/paper/1908.04616/) (2019)

## Built on by (most cited first, 9 of 9)

- [Concerto: Joint 2D-3D Self-Supervised Learning Emerges Spatial Representations](https://mlsanity.com/paper/2510.23607/) (2025)
- [Feat2GS: Probing Visual Foundation Models with Gaussian Splatting](https://mlsanity.com/paper/2412.09606/) (2024)
- [DepthCues: Evaluating Monocular Depth Perception in Large Vision Models](https://mlsanity.com/paper/2411.17385/) (2024)
- [CF3: Compact and Fast 3D Feature Fields](https://mlsanity.com/paper/2508.05254/) (2025)
- [EgoDTM: Towards 3D-Aware Egocentric Video-Language Pretraining](https://mlsanity.com/paper/2503.15470/) (2025)
- [Data-Efficient Inference of Neural Fluid Fields via SciML Foundation Model](https://mlsanity.com/paper/2412.13897/) (2024)
- [Masked Scene Modeling: Narrowing the Gap Between Supervised and Self-Supervised Learning in 3D Scene Understanding](https://mlsanity.com/paper/2504.06719/) (2025)
- [Forging Spatial Intelligence: A Roadmap of Multi-Modal Data Pre-Training for Autonomous Systems](https://mlsanity.com/paper/2512.24385/) (2025)
- [LiDAR-Anchored Collaborative Distillation for Robust 2D Representations](https://mlsanity.com/paper/2602.12524/) (2026)

## Abstract (arXiv)

To advance the state of the art in the creation of 3D foundation models, this paper introduces the ConDense framework for 3D pre-training utilizing existing pre-trained 2D networks and large-scale multi-view datasets. We propose a novel 2D-3D joint training scheme to extract co-embedded 2D and 3D features in an end-to-end pipeline, where 2D-3D feature consistency is enforced through a volume rendering NeRF-like ray marching process. Using dense per pixel features we are able to 1) directly distill the learned priors from 2D models to 3D models and create useful 3D backbones, 2) extract more consistent and less noisy 2D features, 3) formulate a consistent embedding space where 2D, 3D, and other modalities of data (e.g., natural language prompts) can be jointly queried. Furthermore, besides dense features, ConDense can be trained to extract sparse features (e.g., key points), also with 2D-3D consistency -- condensing 3D NeRF representations into compact sets of decorated key points. We demonstrate that our pre-trained model provides good initialization for various 3D tasks including 3D classification and segmentation, outperforming other 3D pre-training methods by a significant margin. It also enables, by exploiting our sparse features, additional useful downstream tasks, such as matching 2D images to 3D scenes, detecting duplicate 3D scenes, and querying a repository of 3D scenes through natural language -- all quite efficiently and without any per-scene fine-tuning.