learnopencv/DUSt3R-Dense-3D-Reconstruction

DUSt3R: Geometric 3D Vision Made Easy : Explanation and Results

This folder contains the Jupyter Notebooks and Scripts for the LearnOpenCV article - DUSt3R: Dense and Unconstrained Stereo 3D Reconstruction.

subsetsTo run:

Visit DUSt3R Repository : Link

git clone --recursive https://github.com/naver/dust3r
cd dust3r
# if you have already cloned dust3r:
# git submodule update --init --recursive

#setup
!pip install -r requirements.txt
# Optional: you can also install additional packages to:
# - add support for HEIC images
# - add pyrender, used to render depthmap in some datasets preprocessing
# - add required packages for visloc.py
!pip install -r requirements_optional.txt

Download checkpoints: mkdir -p checkpoints/ !wget https://download.europe.naverlabs.com/ComputerVision/DUSt3R/DUSt3R_ViTLarge_BaseDecoder_512_dpt.pth -P checkpoints/

Gradio Demo

The input can be a single image or multiple images, and if you have limited GPU Resources say 6GB or 12 GB, for larger scenes go with one-ref pairing strategy. Otherwise use complete-pairing strategy for small subsets.

---

Datasets used in Article:

1. Tanks and Temples - Family | Church >>> Image Set
2. CO3D - motorcycle | teddybear
3. DSLR Camera: Download Link

Download individual Point Cloud Outputs of DUSt3R if you are limited with internet pack:

1. Teddy Subset - DUSt3R GLB - oneref
2. DSLR Scene - DUst3R GLB - oneref
3. Church Subset - DUSt3R GLB - oneref
4. Bike - DUSt3R GLB - oneref

To download all the dataset scenes and point cloud in one go hit the below button:

---

AI Courses by OpenCV

Want to become an expert in AI? AI Courses by OpenCV is a great place to start.