AI Paper Atlas

Research map / 3D vision / 3D generation

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

Xiaoshuai Zhang, Zhicheng Wang, Howard Zhou, Soham Ghosh, Danushen L Gnanapragasam, Varun Jampani, Hao Su, Leonidas Guibas

ECCV 2024 · 3D vision › 3D generation · approach: Learning 3D reconstruction from images · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
unfreeze the 2D encoder during joint training (ablation)ConDense · ScanObjectNN / ImageNet-1kaccuracy: 91.3 -> 94.1 (3D); 86.5 -> 89.6 (2D)Improved
add 2D-3D key point loss L_p (ablation)ConDense · ScanObjectNN / ImageNet-1kaccuracy: 90.5 -> 94.1 (3D); 87.1 -> 89.6 (2D)Improved
add 2D fidelity loss to DINOv2 features (ablation)ConDense · ScanObjectNN / ImageNet-1kaccuracy: 89.7 -> 94.1 (3D); 79.9 -> 89.6 (2D)Improved
use full pre-training data instead of 10% (ablation)ConDense · ScanObjectNN / ImageNet-1kaccuracy: 88.7 -> 94.1 (3D); 79.1 -> 89.6 (2D)Improved
2D-3D consensus pre-training of DINOv2 backboneDINOv2 ViT-g/14 · ImageNet-1k, Places205, ADE20k, PascalVOCaccuracy / mIoU: ImageNet 86.5 -> 89.6; ADE20k 49.0 -> 53.6Improved
use higher-quality NeRFs (4k steps instead of 2k) for pre-training (ablation)ConDense (lighter version) · ImageNet / ScanObjectNNaccuracy: 86.2 / 88.9 -> 87.7 / 93.2Improved
use MinkowskiNet instead of PointBERT as 3D backbone (ablation)ConDense · ScanObjectNNaccuracy: 87.2 -> 93.2Improved
use key point matching instead of global features for scene queries (ablation)ConDense · ScanNet scene retrievalaccuracy: 70.1 -> 78.4Improved
apply 2D-3D consensus to DINOv2 features for depth estimationConDense · NYUd depth (lin. 4)RMSE: 0.298 -> 0.322Hurt
pre-train with ConDense when 3D labeled data is scarce3D segmentation · ScanNet 1%mIoU: Swin3D 54.8 -> ConDense 67.3Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy