Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Scene reconstruction with moving objects
The ability to perceive and understand the static environment (indoor room, street view, etc.) with moving objects (human, car, etc.) from an image sequence is an essential task in 3D vision. We have also witnessed many great processes in both 1) the static scene reconstruction over the past decade years [1,2] and 2) object level detection, segmentation, representation, pose estimation and reconstruction. However, how to efficiently obtain and jointly optimize the full 4D scene including the static and moving parts still has not been well-explored.
This project attempts to reconstruct the geometric and appearance of 4D scenes (static scene + moving objects) [3]. We will start with decomposable radiance field reconstruction with a specific setting: a middle scale static environment (room or outdoor street) and one class of objects (human or car).
Keywords: 3D reconstruction; 3D representation
This project attempts to reconstruct the geometric and appearance of 4D scenes (static scene + moving objects) [3]. We will start with decomposable radiance field reconstruction with a specific setting: a middle scale static environment (room or outdoor street) and one class of objects (human or car).
Task:
Paper, data, code review, running baseline on static scene reconstruction, 3D pose estimation.
Per-category object modeling with tensorial decomposition [4].
[1] https://grail.cs.washington.edu/rome/
[2] Sun et al. "NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video." CVPR 2021
[3] Abhijit et al. "Panoptic Neural Fields: A Semantic Object-Aware Neural Scene Representation." CVPR 2022
[4] Chen and Xu et al. "TensoRF: Tensorial Radiance Fields." ECCV 2022
Requirement:
3D Vision; PyTorch; Machine learning and deep learning
This project attempts to reconstruct the geometric and appearance of 4D scenes (static scene + moving objects) [3]. We will start with decomposable radiance field reconstruction with a specific setting: a middle scale static environment (room or outdoor street) and one class of objects (human or car).
Task:
Paper, data, code review, running baseline on static scene reconstruction, 3D pose estimation.
Per-category object modeling with tensorial decomposition [4].
[1] https://grail.cs.washington.edu/rome/
[2] Sun et al. "NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video." CVPR 2021
[3] Abhijit et al. "Panoptic Neural Fields: A Semantic Object-Aware Neural Scene Representation." CVPR 2022
[4] Chen and Xu et al. "TensoRF: Tensorial Radiance Fields." ECCV 2022
Requirement: 3D Vision; PyTorch; Machine learning and deep learning