源码地址
代码地址:https://github.com/castacks/tartanvo/tree/python3
配环境
git clone https://github.com/castacks/tartanvo.git -b python3
创建conda环境:
conda create -n tartanvo python=3.8
安装pytorch
conda install pytorch==1.10.1 torchvision==0.11.2 torchaudio==0.10.1 cupy cudatoolkit=11.3 -c pytorch -c conda-forge
安装其他库
python3 -m pip install pybind11==2.3
pip install scipy==1.4
pip install numpy==1.18
pip install matplotlib==3.4.3
pip install opencv-python==4.2.0.32 -i https://pypi.tuna.tsinghua.edu.cn/simple
准备数据
下载模型
mkdir models
wget https://cmu.box.com/shared/static/t1a5u4x6dxohl89104dyrsiz42mvq2sz.pkl -O models/tartanvo_1914.pkl
下载测试数据
mkdir data
wget https://cmu.box.com/shared/static/nw3bi7x5vng2xy296ndxt19uozpk64jq.zip -O data/KITTI_10.zip
unzip -q data/KITTI_10.zip -d data/KITTI_10/
测试
KITTI 数据集上进场测试,创建run.sh
python3 vo_trajectory_from_folder.py --model-name tartanvo_1914.pkl \
--kitti \
--batch-size 1 --worker-num 1 \
--test-dir data/KITTI_10/image_left \
--pose-file data/KITTI_10/pose_left.txt
跑起来了,哈哈哈