Unifying Fine-grained Perception into MLLMs w/o Task Decoders. 16 tokens enable precise segmentation

October 1, 2025 · View on GitHub

This repo is the official implementation of paper: 🛸 UFO: A Unified Approach to Fine-grained Visual Perception via Open-ended Language Interface as well as the follow-ups. We have made every effort to ensure that the codebase is clean, concise, easily readable, state-of-the-art, and relies only on minimal dependencies.

UFO: A Unified Approach to Fine-grained Visual Perception via Open-ended Language Interface

Hao Tang, Chenwei Xie , Haiyang Wang, Xiaoyi Bao, Tingyu Weng, Pandeng Li, Yun Zheng $^\dagger$ , Liwei Wang $^\dagger$

Primary contact: Hao Tang ( tanghao@stu.pku.edu.cn )

📣 News

[25-10-1] We release checkpoints of UFO-InternVL2.5-8B in repo.
[25-9-19] 🔥 UFO is accepted by NeurIPS 2025 as a Spotlight!
[25-3-12] We release separate repos of UFO-InternVL2-8B and add REC inference on InternVL repo.
[25-3-4] 🚀 Training and inference Code is released.
[25-3-3] 👀 UFO is released on arXiv.

Overview

👀 Todo

Release the arXiv version.
Release code and models of multi-task training on UFO-ViT.
Release code and models of fine-grained instruction tuning on UFO-InternVL2.5-8B and UFO-LLaVA-1.5-7B.
Release full code and models of multi-task training on UFO-InternVL2.5-8B.

🤔 Introduction

Previous efforts to introduce fine-grained perception tasks into MLLMs rely heavily on task-specific decoders or suboptimal formats (e.g., polygons), impeding the visual unified modeling. To overcome this, we propose UFO:

😮 We reformulate segmentation as embedding retrieval, where the mask token embedding computes similarity with image features by dot product, retrieving high-similarity positions to generate the mask.
🚀 We first explore the image representation capabilities of MLLMs. We argue that since MLLMs excel in understanding, the mask information is also in the image features and we just need to retrieve it.
🤗 Fully aligned with open-ended Language interface: UFO unifies detection and segmentation through the open-ended language interface without any additional decoders, enabling seamless integration with MLLMs.
🔥 Competitive performance: UFO surpasses GiT, a text-based generalist model, by 12.3 mAP on COCO instance segmentation and 3.3 mIoU on ADE20K. It also matches or exceeds decoder-based methods in various grounding tasks, eliminating the need for task-specific decoders.

🚀 Main Results

Single-Task Benchmark

Model	Params	Metric	Perfomance	ckpt	config
UFO-ViT-B_detection	131M	mAP	47.8	ckpt	config
UFO-ViT-B_insseg	131M	mAP	42.6	ckpt	config
UFO-ViT-B_semseg	131M	mIoU	49.5	ckpt	config
UFO-ViT-B_caption	131M	BLEU-4	34.2	ckpt	config
UFO-ViT-B_grounding	131M	Acc@0.5	83.6	ckpt	config

Multi-Task Benchmark

Model	Params	Detection	Ins Seg	Sem Seg	Caption	Grounding	ckpt	config
UFO-ViT-B_multi-task	131M	48.3	43.5	50.2	35.3	85.8	ckpt	config
UFO-ViT-L_multi-task	387M	52.9	47.3	54.0	35.9	88.5	ckpt	config
UFO-ViT-H_multi-task	756M	54.1	48.1	55.7	37.6	89.2	ckpt	config

Task Synergy in Multi-Tasking Training

Model	Params	Detection	Ins Seg	Sem Seg	Caption	Grounding
UFO-B_single-task	131M	47.8	42.6	49.5	34.2	83.6
Improvement		+0.5	+0.9	+0.7	+1.1	+2.2
UFO-B_multi-task	131M	48.3	43.5	50.2	35.3	85.8

MLLM Performance on Multi-Task Benchmark

UFO-InternVL2.5-8B:

Resolution	Detection	Ins Seg	Sem Seg	Caption	Grounding	ckpt	config
448x448	44.0	37.4	53.9	39.6	90.4	ckpt	config
896x896	50.9	43.6	54.6	-	-	ckpt	config
1344x1344	51.9	45.2	-	-	-	ckpt	config

Visual Grounding

RefCOCO Validation Set

Model	REC	RES	ckpt	config
UFO-LLaVA-1.5-7B	89.9	76.2	ckpt	config
UFO-LLaVA-1.5-7B (ft)	90.8	77.2	ckpt	config
UFO-InternVL2.5-8B	91.8	80.0	ckpt	config
UFO-InternVL2.5-8B (ft)	93.1	81.0	ckpt	config

Reasoning Segmentation

Model	Overall	Short Query	Long Query	ckpt	config
UFO-LLaVA-1.5-7B	53.8	40.1	58.2	ckpt	config
UFO-LLaVA-1.5-7B (ft)	58.0	46.3	61.7	ckpt	config
UFO-InternVL2.5-8B	60.0	48.7	63.6	ckpt	config
UFO-InternVL2.5-8B (ft)	67.0	56.2	70.4	ckpt	config

🛠️ Quick Start

Installation

conda create -n UFO python=3.11

conda activate UFO

pip install torch==2.1.0+cu118 torchvision==0.16.0+cu118 -f https://download.pytorch.org/whl/torch_stable.html
pip install -U openmim
mim install "mmengine==0.8.3"
mim install "mmcv==2.1.0"
pip install "transformers==4.37.2"

git clone git@github.com:nnnth/UFO.git
cd UFO

pip install -v -e .
pip install -r requirements/optional.txt
pip install -r requirements/runtime.txt

(Optional) Install Java manually for image caption evaluation. Without Java, you can train image caption normally, but fail in caption evaluation.

Dataset Preparation

Multi-Tasking Dataset

We follow GiT to prepare the multi-task datasets. Please refer here for more details.

Instruction Tuning Dataset

We use 24 datasets in for instruction tuning on MLLMs. For more details, please refer here.

Download Pretraining Weight

We use LLaVA-1.5-7B and InternVL2.5-8B as MLLM pretraining. In multi-task training on UFO-ViT, we also use Bert Tokenizer and Bert Embeddings. Please download and organize them as follows:

UFO
|──ckpt
|──|──llava-1.5-7b-hf
|──|──InternVL2_5-8B
|──|──bert-base-uncased
|──|──bert_embed_womask.pt
|──|──bert_embed.pt
|──|──bert_embed_large.pt
|──|──bert_embed_huge.pt

For InternVL2_5-8B, we add a custom function for lora training. Please replace the original file following issue.

Demo

Please download checkpoints from kanashi6/UFO, then save them under root dir:

UFO
|──ufo-vit-b-single-det.pth
|──ufo-vit-b-single-insseg.pth
|──...

Run demo on detection (coco):

python demo.py --img_path demo/demo.jpg --config configs/UFO-ViT/single_detection_base.py \
  --ckpt_path ./ufo-vit-b-single-det.pth --out_dir ./vis/ --task detection

Run demo on RES:

python demo.py --img_path demo/demo.jpg --config configs/UFO-InternVL2_5-8B/internvl2_5_8b_res_ft_2w.py \
  --ckpt_path ./ufo-internvl2_5-8b-res.pth --out_dir ./vis/ --task res --text bench

Scripts

For Training and evaluation commands, please refer here.

👍 Acknowledgement

MMDetection The codebase we built upon. Thanks for providing such a convenient framework.
GiT We use the multi-task benchmark established by GiT.
InternVL We borrow codes of MLLMs from InternVL repo.

📘 Citation

Please consider citing our work as follows if it is helpful.

@article{tang2025ufo,
    title={UFO: A Unified Approach to Fine-grained Visual Perception via Open-ended Language Interface},
    author={ Hao Tang, Chenwei Xie, Haiyang Wang, Xiaoyi Bao, Tingyu Weng, Pandeng Li, Yun Zheng, Liwei Wang},
    journal={arXiV:2503.01342},
    year={2025}
}