2. 📦 Environment Setup

Create Conda Environment

conda create -n scarf-slam python=3.11
conda activate scarf-slam

Install Depth Anything 3

Check the official repository for more details.

cd ~/git/
git clone https://github.com/ByteDance-Seed/Depth-Anything-3.git
cd Depth-Anything-3

pip install xformers torch\>=2 torchvision
pip install -e .

(Optional) To include sky masks in DA3 predictions, open da3.py, find the NestedDepthAnything3Net class, then update its forward function by adding the following line just before the return statement:

output.sky = metric_output.sky

Install Other Dependencies

pip install rosbags open3d gtsam vismatch

• rosbags: pure-Python library for reading and writing ROS bag files.
• open3d: library for 3D data processing and visualization.
• gtsam: factor-graph optimization library for robotics and computer vision.
• vismatch: unified API for image matching models.

(Optional) Install OV-SLAM

OV-SLAM refers to OpenVINS plus ov_secondary for pose graph optimization.

sudo apt update
sudo apt install libeigen3-dev libboost-all-dev libceres-dev

cd ~/ros2_ws/src
git clone https://github.com/rpng/open_vins.git
git clone https://github.com/ori-drs/ov_secondary_scarf.git
cd ~/ros2_ws
colcon build --symlink-install

Our implementation requires Ceres 2.2.0. For compatibility with older Ceres versions, use the compat/ceres-2.0.0 branch.