|
--- |
|
license: mit |
|
pipeline_tag: image-to-3d |
|
language: |
|
- en |
|
--- |
|
## DreamCar: Leveraging Car-specific Prior for in-the-wild 3D Car Reconstruction |
|
|
|
|
|
![image/png](https://cdn-uploads.huggingface.co/production/uploads/64ad086c5d48838462e2eee1/2MWeSncAOUfhvAw0-VFHz.png) |
|
|
|
## ๐ฎ Highlights |
|
|
|
DreamCar can reconstruct a completed 3D car model in the moving-forward scene given a few images even a single image. |
|
|
|
### ๐ฅ Symmetry Learning, Car-specific Piror, Pose Optimization |
|
- Symmetry Learning: We use the symmetric nature of cars to reconstruct a completed 3D car. |
|
- Car generative prior: We contribute a 3D car dataset to empower Zero123 with car-specific piror. |
|
- Pose optimization: We propose a pose optimization method to optimize the poses with error from self-driving datasets. |
|
|
|
|
|
## ๐ฉ **Updates** |
|
|
|
Welcome to **watch** ๐ this repository for the latest updates. |
|
|
|
โ
**[2024.7.23]** : Release [project page](https://xiaobiaodu.github.io/dreamcar-project/). |
|
|
|
โ
**[2024.7.23]** : Code Release. |
|
|
|
|
|
You can find our code in https://github.com/xiaobiaodu/DreamCar |
|
|
|
|
|
|
|
|
|
|
|
|
|
|