Code for the baseline experiments on Conflab-dataset

Prerequisites

• PyTorch 1.7+
• detectron2

Install the required packages by:

pip install requirements.txt

or use uv. The packages were added to lock file like so

uv add scikit-learn seaborn parse rich tqdm opencv-python detectron2 torch==1.10.1+cpu torchvision==0.11.2+cpu \
--prerelease allow \
-f https://dl.fbaipublicfiles.com/detectron2/wheels/cpu/torch1.10/index.html \
-f https://download.pytorch.org/whl/cpu/torch_stable.html

See the respective websites for the GPU with CUDA packages.

How to Run

Dataset

create dataset by:

python data_create.py create_coco=true force_register=true

Training

bash scripts/script_train.sh -b "R50_FPN" -m train

Evaluation

bash scripts/script_train.sh -b "R50_FPN" -m test