Setting the environment for foundation pose
# Pull the following docker image and launch it:
docker pull shingarey/foundationpose_custom_cuda121:latest
# additional requirements
pip install Hydra fastapi uvicorn
Build FoundationPose
# Enter the container, git clone this repo:
git clone https://github.com/teal024/FoundationPose-plus-plus
# Export project root as the dir of the repo
export PROJECT_ROOT=/root/FoundationPose-plus-plus
# Download FoundationPose weights to $PROJECT_ROOT/FoundationPose/weights
from Google Drive: https://drive.google.com/drive/folders/1DFezOAD0oD1BblsXVxqDsl8fj0qzB82i
# Start building process
cd $PROJECT_ROOT/FoundationPose
bash build_all.sh
Download Other Weights
1. Download Qwen2-VL weights to $PROJECT_ROOT/Qwen2-VL/weights
from huggingface: https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct
2. Download Sam-HQ weights to $PROJECT_ROOT/sam-hq/pretrained_checkpoints
from Google Drive: https://drive.google.com/file/d/1Uk17tDKX1YAKas5knI4y9ZJCo0lRVL0G/view
Utilities
# For SAM-HQ
pip install segment-anything-hq
cd $PROJECT_ROOT/sam-hq
pip install -e .
# For Qwen-VL-2-Instruct
pip install git+https://github.com/huggingface/transformers@21fac7abba2a37fae86106f87fcf9974fd1e3830 accelerate
pip install qwen-vl-utils
# For Cutie
cd $PROJECT_ROOT/Cutie
pip install -e .
python cutie/utils/download_models.py