[Literature Review] General Scene Multi-view 3D reconstruction (3D 복원 기술 관련 기술 정리)

반응형

Category:

• 1. Multiview-geometry
• 2. Multiview-stereo
• 3. Neural reconstruction
• 4. Direct regression

 

1. Multi-view geometry (e.g., Structure from motion).

GLOMAP: Global Structure-from-Motion Revisited [ECCV 2024] 

FrozenRecon: Pose-free 3D Scene Reconstruction with Frozen Depth Models [ICCV 2023]

A Pose-only Solution to Visual Reconstruction and Navigation [TPAMI 2021]

Pixel-Perfect Structure-from-Motion with Featuremetric Refinement [ICCV 2021]

Reducing Drift in Structure From Motion Using Extended Features [3DV 202] 

HSfM: Hybrid Structure-from-Motion [CVPR 2017]

COLMAP: Structure-from-Motion Revisited [CVPR 2016] 

 

A Global Linear Method for Camera Pose Registration [ICCV 2013]

SfM with MRFs: Discrete-Continuous Optimization for Large-Scale Structure from Motion [TPAMPI 2012]

Towards linear-time incremental structure from motion [3DV 2013]

Hybridstructure-from-motion [CVPR 2012]

 

2. Surface reconstruction with Multiview-Stereo (assuming camera poses are given)

GoMVS: Geometrically Consistent Cost Aggregation for Multi-View Stereo [CVPR 2024]

Geomvsnet: Learning multi-view stereo with geometry perception [CVPR 2023]

Constraining depth mapgeometry for multiview stereo: A dual-depth approach with saddle-shaped depth cells [ICCV 2023]

Rethinking Depth Estimation for Multi-View Stereo: A Unified Representation [CVPR 2022]

TransMVSNet: Global Context-aware Multi-view Stereo Network with Transformers [CVPR 2022]

Multiview stereo with cascaded epipolar raft [ECCV 2022]

Mvsnet: Depth inference for unstructured multi-view stereo [ECCV 2018]

Adaptive Surface Reconstruction with Multiscale Convolutional Kernels [ICCV 2021] 

Pixelwise View Selection for Unstructured Multi-View Stereo [ECCV 2016]

 

3. Neural rendering  for 3D reconstruction (assuming camera poses are given)

Neuralangelo: High-Fidelity Neural Surface Reconstruction [CVPR 2023]

Neural 3D Reconstruction in the Wild [SIGGRAPH 2022] 

GenS: Generalizable Neural Surface Reconstruction from Multi-View Images [NeurIPS 2023]

Poco: Point convolution for surface reconstruction [CVPR 2022] 

Neural 3D Scene Reconstruction with the Manhattan-world Assumption [CVPR 2022] 

 

MonoSDF: Exploring Monocular Geometric Cues for Neural Implicit Surface Reconstruction [NeurIPS 2022] 

HF-NeuS: Improved Surface Reconstruction Using High-Frequency Details [NeurIPS 2022]

Geo-Neus: Geometry-Consistent Neural Implicit Surfaces Learning for Multi-view Reconstruction [NeurIPS 2022] 

UNISURF: unifying neural implicit surfaces and radiance f ields for multi-view reconstruction [ICCV 2021]

NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction [NeurIPS 2021]

Atlas: End-to-End 3D Scene Reconstruction from Posed Images [ECCV 2020] 

Multiview neural surface reconstruction by disentangling geometry and appearance [NeurIPS 2020]

 

 

4. Direct image regression for 3D reconstruction (e.g., direct regression of the depth that is under globally coherent scale often with pairwise setup)

Grounding Image Matching in 3D with MASt3R [Arxiv 2024]

DUSt3R: Geometric 3D Vision Made Easy [CVPR 2024]

SimpleRecon 3D Reconstruction without 3D Convolutions [ECCV 2022] 

Towards accurate reconstruction of 3d scene shape from a single monocular image [TPAMI 2022]

Vision Transformers for Dense Prediction [ICCV 2021]

DeepV2D: Video to Depth with Differentiable Structure from Motion [ICLR 2020]

DeepTAM: Deep tracking and mapping with convolutional neural networks [ECCV2018]

GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Pose [CVPR 2018]

DeMoN: Depth and Motion Network for Learning Monocular Stereo [CVPR 2017]

 

 

*Please feel free to leave a reply if you detect any recent papers. :)

*Note: RGB-D based reconstruction methods are not included.

 

- Walking Tree -