TALENT

「CVPR2026」Official implementation of TALENT: Target-aware Efficient Learning for Referring Image Segmentation.

Installation

git clone https://github.com/Kimsure/TALENT.git
cd TALENT

Environment

conda create -n TALENT python=3.9.18 -y
conda activate TALENT
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2
pip install -r requirement.txt

Datasets

The detailed instruction is in prepare_datasets.md

Pretrained weights

Download the pretrained weights of DiNOv2-B, DiNOv2-L and ViT-B to pretrain

mkdir pretrain && cd pretrain
## DiNOv2-B as Image encoder
wget https://dl.fbaipublicfiles.com/dinov2/dinov2_vitb14/dinov2_vitb14_reg4_pretrain.pt
## CLIP ViT-B as Text encoder
wget https://openaipublic.azureedge.net/clip/models/5806e77cd80f8b59890b7e101eabd078d9fb84e6937f9e85e4ecb61988df416f/ViT-B-16.pt

Quick Start

To evaluate TALENT, run the following script (Remember adjusting your own MODEL/DATASET PATH.)

bash run_scripts/test.sh

If you want to visualize the results, simply modify the visualize to True in the config file.

Weights

Our model weights have already been open-sourced and can be directly downloaded from here.

Acknowledgements

The code is based on CRIS, ETRIS and TALENT. We thank the authors for their open-sourced code and encourage users to cite their works when applicable.

Citation

If you find our work useful, please cite this paper: