BodySLAM 專案目不完備，而且確實沒有setup.py或pyproject.toml檔案可執行pip install -e .
我是先將Hugging Face中CycleVO.pth下載

python -m venv venv
source venv/bin/activate  

#查看 src 目錄下的程式碼，找出需要的依賴
pip install torch torchvision numpy opencv-python matplotlib
pip install transformers # 如果使用 Hugging Face 模型
cd ~/BodySLAM
export PYTHONPATH="${PYTHONPATH}:~/BodySLAM/src"
pip install timm==0.6.7

python examples/depth_estimation/basic_depth_estimation.py #得depth_map1.png
python examples/depth_estimation/depth_map_scaling.py #得colorized_depth_map1.png
python examples/depth_estimation/batch_processing.py #得depth_image1~4.png

python examples/pose_estimation/run_cycle_pose.py --mode pair \
    --model_path ~/BodySLAM/models/CycleVO.pth \
    --input ~/BodySLAM/examples/pose_estimation/resources/pair/frameA.jpg \
    --input2 ~/BodySLAM/examples/pose_estimation/resources/pair/frameB.jpg \
    --output pose.txt
    
python examples/pose_estimation/run_cycle_pose.py --mode sequence \
    --model_path ~/BodySLAM/models/CycleVO.pth \
    --input ~/BodySLAM/examples/pose_estimation/sequence \
    --output sequence_poses.txt

前3個結果如它github所附圖;後2個它會給你pose矩陣,最後一個我是將之前MITI簡幾百張圖,沒有網頁展示的3D重建