Official Implementation of paper:

[ICCV2025]Advancing Visual Large Language Model for Multi-granular Versatile Perception

Checkpoint Download

Checkpoints can be downloaded at MVP-LM-checkpoint

Installation

pip install -r requirements.txt
pip install -e /PATH-to-mvp-lm/mvp-lm/psalm/model/mask_decoder/Openseed_Simplify/modeling/pixel_decoder/ops
pip install torch==1.13.1+cu116
pip install torchaudio==0.13.1+cu116
pip install torchvision==0.14.1+cu116

Code-related

The source code of the shell script required to run the code contains many personalized logic elements, aiming to improve efficiency, establish a standard operating procedure (SOP) for each execution, and automate the model training and inference process. You may extract the most essential command for launching the Python process and incorporate your own automation logic as needed.

Aknowledgements

This project is based on the following great works: Mask2Former, PSALM and OpenSeeD.