This repository hosts USC ISI / CORAL teams systems for LWLL evaluations.
Documentation is available at: https://usc-isi.lollllz.com/coral/
This folder is for CORAL TEAM Object Detection System. Currently, this is the Dry Run version.
For building the docker for evaluation:
docker build -t detection_pipeline .
For running the Object Detection Pipeline:
docker run --rm -ti --gpus '"device=0"' --env-file env.list --ipc=host -v /lwll/evaluation:/lwll/evaluation detection_pipeline
Assuming LWLL* environment variables are correctly set. If the above command is used, then it will be read from env.list file.
Assuming all datasets are stored at /lwll/evaluation/.
The system uses only one GPU. Please make sure to give access to only one GPU using --gpus '"device=0"'.