DiffVL

12/08/2023 Release the code of UI-system and DiffVL algorithm.

Prerequsite

Docker

Install Nvidia-docker. Then:

cd DiffVL/docker
sudo bash build.sh
sudo bash run.sh
sudo bash start.sh

Conda (Linux only)

conda create -n anno python
conda activate anno
cd TaskAnnotator-cf
pip install -e .
pip install "git+https://github.com/facebookresearch/pytorch3d.git"

Test the environment:

For UI system:

python ./ui/tester/test_viewer.py 0
flask --app ./frontend/flaskr init-db

For DiffVL algorithm:

python diffsolver/main.py

Download Dataset:

Download the Dataset on the google drive. Put the dataset in the folder: DiffVL/diffsolver/assets/

Make sure the path of every single data follows the format: DiffVL/Diffsolver/assets/Task/Task/data/task_x

Run:

UI-system:

flask --app ./frontend/flaskr init-db
flask --app ./frontend/flaskr --debug run --host 0.0.0.0

Then open http://127.0.0.1:5000 in your browser to see the annotator.

DiffVL solver:

Single stage Tasks

python diffsolver/run_single_stage.py lang --config diffsolver/examples/singlestage_dev/task2_wind.yaml 

You could choose the config YAML file in the examples/ folder.

Multistage Tasks

python diffsolver/run_multistage.py lang --config diffsolver/examples/multistage/task1/total.yml

You could choose the config YAML file in the examples/ folder.

Cite:

If you find this codebase useful in your research, please consider citing: