The source code of our paper: LGR-NET: Language Guided Reasoning Network for Referring Expression Comprehension

python==3.7.13
torch==1.11.0+cu113
torchvision==0.12.0+cu113
mmcv-full==1.3.18
mmdet==2.11.0
tensorboard==2.8.0
transformers==2.11.0
einops==0.4.1
icecream==2.1.2
numpy==1.22.3
scipy==1.8.0
ftfy==6.1.1

We recommand to install mmdet from the source code in this repository (./models/swin_model)

We follow the data preparation of TransVG, which can be found in GETTING_STARTED.md

We use the checkpoints from QRNet, they put the checkpoints in Google Drive. You should mkdir checkpoints/ and put the downloaded checkpoints in it.

bash train.sh

bash test.sh

This code is partially based on TransVG and QRNet