Yufei Liu's Homepages

Example Talk

Sat, 01 Jun 2030 13:00:00 +0000

Click on the Slides button above to view the built-in slides feature.

Slides can be added in a few ways:

Create slides using Hugo Blox Builder’s Slides feature and link using slides parameter in the front matter of the talk file
Upload an existing slide deck to static/ and link using url_slides parameter in the front matter of the talk file
Embed your slides (e.g. Google Slides) or presentation video on this page using shortcodes.

Further event details, including page elements such as image galleries, can be added to the body of this page.

FLUX.2-based Image Identity Consistency Enhancement

Thu, 01 Oct 2026 00:00:00 +0000

Developed during an algorithm internship with the face team at Meitu MT-Lab, this project addresses identity drift in AIGC portrait editing. The module takes an edited portrait and a user reference image as inputs, then refines identity while retaining the intended edit.

I built the data pipeline for face detection, identity clustering, mask generation, canonical alignment, reference matching, and quality filtering. I also designed joint perceptual, identity, and contour objectives for multi-step FLUX.2 LoRA training and developed evaluation tools for identity similarity, facial contour change, and denoising trajectories.

The resulting model improved reference identity similarity across internal test sets and was deployed in the telephoto portrait feature of AI Camera.

Qualitative Results

Frontal Face Reshaping × JJ Lin

Identity Similarity +0.186

Input

Reference

Output

Input–Reference 0.410　→　Output–Reference 0.596

Half-Profile Face Reshaping × Jackson Yee

Identity Similarity +0.298

Input

Reference

Output

Input–Reference 0.561　→　Output–Reference 0.859

Strong Stage Lighting × G-Dragon

Identity Similarity +0.389

Input

Reference

Output

Input–Reference 0.286　→　Output–Reference 0.675

Overall Evaluation

0.406

0.680

Celeb-new

+0.274

115 cases, 100.0% improved
Median pose gap 1.8°, non-face MAE 0.060

0.272

0.600

Celeb-old

+0.328

28 cases, 96.4% improved
Median pose gap 1.4°, non-face MAE 0.043

0.338

0.721

Colleague

+0.383

440 cases, 99.8% improved
Median pose gap 3.1°, non-face MAE 0.227

Input → ReferenceOutput → Reference

Identity similarity improves consistently across all three test sets, while pose changes and non-face-region differences remain limited.

SynPo: Boosting Training-Free Few-Shot Medical Segmentation via High-Quality Negative Prompts

Tue, 01 Sep 2026 00:00:00 +0000

SynPo addresses weak negative prompts in training-free few-shot medical image segmentation. It combines SAM and DINOv2 features to construct a synergistic confidence map, then selects informative positive and negative points through confidence-aware sampling and clustering.

Method	CHAOS-MRI					Synapse-CT
Method	Spleen	Liver	LK	RK	Mean	Spleen	Liver	LK	RK	Mean
SSL-ALPNettrained	67.02	73.05	73.63	78.39	73.02	60.25	73.65	63.34	54.82	63.02
ADNettrained	75.92	80.81	75.28	83.28	78.82	63.48	77.24	72.13	79.06	72.97
Q-Nettrained	75.99	81.74	78.36	87.98	81.02	74.86	71.21	75.26	74.79	74.03
RPTtrained	76.37	82.86	80.72	89.82	82.44	79.13	82.57	77.05	72.58	77.83
GMRDtrained	76.09	81.42	83.96	90.12	82.90	78.31	79.60	81.70	74.46	78.52
PerSAMtraining-free	69.14	42.44	64.84	71.36	61.12	65.03	65.55	58.47	60.31	62.34
ProtoSAMtraining-free	76.51	81.94	71.46	81.43	77.83	65.50	87.84	69.44	71.04	73.45
SynPo (ours)training-free	80.30	77.32	77.32	83.04	79.50	83.76	81.32	75.00	79.63	79.91

Qualitative results on abdominal MRI. Colored overlays show the predicted organ masks.

Module ablation

SAM	DINO	PSM	NRM	Dice	Gain
✓				57.48	–
✓	✓			66.70	+9.22
✓	✓	✓		78.74	+12.04
✓	✓	✓	✓	79.93	+1.19

Negative-prompt strategy on CHAOS

Model	Dice
PerSAM	56.72
+ Negative Point Selection	68.90

SynPo achieves a mean Dice score of 79.91 on Synapse-CT without task-specific training. The module study shows the largest gain comes from the Point Selection Module.

OmniPersona: Holistic Identity-Preserving Video Creation and Editing via Spatiotemporal Decoupled Persona Injection in Unified DiT

Wed, 01 Jul 2026 00:00:00 +0000

Role: Co-first Author (Equal Contribution)

Overview

Generation Example

The static input conditions and the animated output are combined into one synchronized video, following the presentation layout.

OmniPersona introduces Multimodal Holistic Persona Representation and spatiotemporally decoupled persona injection for unified video creation and editing. I co-led the work and contributed to the holistic identity representation module, combining face identity features with DINOv2 and CLIP features for clothing, hairstyle, and body characteristics.

On reference-to-video generation, OmniPersona reaches 79.3% face similarity and 68.7% body similarity, improving over the strongest baseline by 5.2 and 34.2 percentage points, respectively.

OneFont: A Unified Agent for End-to-End Font Creation

Sun, 01 Mar 2026 00:00:00 +0000

From Manual Model Selection to OneFont

OneFont completes model selection and refinement in a single agent workflow.

Motivation

Existing diffusion and MLLM-based font generation systems often focus on a single editing task. Supporting a new script or generation workflow can require substantial task-specific fine-tuning, while selecting the right model and repairing an imperfect result still depend on expert intervention.

OneFont treats end-to-end font creation as an MLLM agent task. Given a user request, it selects suitable models and tools, generates candidate results, verifies readability and style, and performs local refinement when needed.

Method Overview

OneFont converts each request into a reasoning trace and structured tool calls. Its generation library covers text-to-art fonts, style transfer, vector font generation, retrieval, handwriting generation, and local editing. Supervised fine-tuning teaches the agent to plan and invoke these tools, while GRPO-based preference alignment improves the quality of its generation strategy. During inference, a graph-based planner verifies intermediate results, backtracks when necessary, and applies local refinement.

Training framework and graph-based inference planner

Tool LibraryText, image, and answer tools for multimodal font creation

Model TrainingSFT builds tool-use ability, then GRPO aligns generation preferences

Model InferenceThe planner verifies results, backtracks when needed, and refines the output

Qualitative Results

Across text-rich generation tasks, OneFont produces readable text while matching the requested object, composition, and visual style. The comparison includes general image generators and dedicated text rendering models.

Qualitative comparison on cakes, T-shirts, book covers, and character images with embedded text. The final column shows OneFont results.

Wan-VACE-based Video Object and Shadow Removal

Thu, 01 Jan 2026 00:00:00 +0000

This Huawei collaboration project addresses the shortage of high-quality paired data for video object removal, especially when shadows and reflections must also be removed.

I independently built a data generation and fine-tuning pipeline using Wan2.2-5B, SAM2, and a baseline removal model. From roughly 4,000 generated groups, I selected 1,070 high-quality training pairs and retained 2,500 difficult cases for evaluation. I then fine-tuned the VACE module of Wan2.1-VACE-1.3B with DiffSynth-Studio, Accelerate, and DeepSpeed on six NVIDIA A800 GPUs.

All videos are displayed at 480 × 276, matching the output resolution of our model.

Condition Video vs. Baseline

Condition Video Baseline ↔

Drag the slider to compare both videos frame by frame.

Condition Video vs. Ours

Condition Video Ours ↔

Drag the slider to inspect object and shadow removal.

Human evaluation found 2,446 of the 2,500 difficult cases usable, a 97.84% success rate, with improved object-shadow removal and background reconstruction.

Clinical Study on the Application of a High-Sensitivity Electronic Nose on Thin-Film Gas Sensor Array Technology Combined with Deep Learning Algorithm for Early Non-Invasive Diagnosis of Chronic Atrophic Gastritis

Mon, 01 Sep 2025 00:00:00 +0000

This study evaluates breath analysis with a thin-film gas sensor array and learning-based models for early, non-invasive diagnosis of chronic atrophic gastritis.

TEmory: A Temporal-Memory Approach to Weakly Supervised Colonic Polyp Frame Detection

Fri, 01 Nov 2024 00:00:00 +0000

TEmory models relationships between adjacent video frames and stores representative normal and polyp features in parallel memory banks, improving robustness to minority normal structures in colonoscopy videos.

Projects

Sun, 19 May 2024 00:00:00 +0000

An Interpretable Approach to the Solutions of High-Dimensional Partial Differential Equations

Thu, 01 Feb 2024 00:00:00 +0000

The method transfers the structure of one-dimensional analytical solutions to higher-dimensional forms, then uses genetic programming and automatic differentiation to search for mathematically interpretable solutions that satisfy the governing equations and boundary conditions.

Learn JavaScript

Tue, 24 Oct 2023 00:00:00 +0000

Hugo Blox Builder is designed to give technical content creators a seamless experience. You can focus on the content and the Hugo Blox Builder which this template is built upon handles the rest.

Embed videos, podcasts, code, LaTeX math, and even test students!

On this page, you’ll find some examples of the types of technical content that can be rendered with Hugo Blox.

Video

Teach your course by sharing videos with your students. Choose from one of the following approaches:

Youtube:

{{< youtube w7Ft2ymGmfc >}}

Bilibili:

{{< bilibili id="BV1WV4y1r7DF" >}}

Video file

Videos may be added to a page by either placing them in your assets/media/ media library or in your page’s folder, and then embedding them with the video shortcode:

{{< video src="my_video.mp4" controls="yes" >}}

Podcast

You can add a podcast or music to a page by placing the MP3 file in the page’s folder or the media library folder and then embedding the audio on your page with the audio shortcode:

{{< audio src="ambient-piano.mp3" >}}

Try it out:

Test students

Provide a simple yet fun self-assessment by revealing the solutions to challenges with the spoiler shortcode:

{{< spoiler text="👉 Click to view the solution" >}}
You found me!
{{< /spoiler >}}

renders as

👉 Click to view the solution

You found me 🎉

Math

Hugo Blox Builder supports a Markdown extension for $\LaTeX$ math. You can enable this feature by toggling the math option in your config/_default/params.yaml file.

To render inline or block math, wrap your LaTeX math with {{< math >}}$...${{< /math >}} or {{< math >}}$$...$${{< /math >}}, respectively.

We wrap the LaTeX math in the Hugo Blox math shortcode to prevent Hugo rendering our math as Markdown.

Example math block:

{{< math >}}
$$
\gamma_{n} = \frac{ \left | \left (\mathbf x_{n} - \mathbf x_{n-1} \right )^T \left [\nabla F (\mathbf x_{n}) - \nabla F (\mathbf x_{n-1}) \right ] \right |}{\left \|\nabla F(\mathbf{x}_{n}) - \nabla F(\mathbf{x}_{n-1}) \right \|^2}
$$
{{< /math >}}

renders as

$$\gamma_{n} = \frac{ \left | \left (\mathbf x_{n} - \mathbf x_{n-1} \right )^T \left [\nabla F (\mathbf x_{n}) - \nabla F (\mathbf x_{n-1}) \right ] \right |}{\left \|\nabla F(\mathbf{x}_{n}) - \nabla F(\mathbf{x}_{n-1}) \right \|^2}$$

Example inline math {{< math >}}$\nabla F(\mathbf{x}_{n})${{< /math >}} renders as $\nabla F(\mathbf{x}_{n})$ .

Example multi-line math using the math linebreak (\\):

{{< math >}}
$$f(k;p_{0}^{*}) = \begin{cases}p_{0}^{*} & \text{if }k=1, \\
1-p_{0}^{*} & \text{if }k=0.\end{cases}$$
{{< /math >}}

renders as

$$ f(k;p_{0}^{*}) = \begin{cases}p_{0}^{*} & \text{if }k=1, \\ 1-p_{0}^{*} & \text{if }k=0.\end{cases} $$

Code

Hugo Blox Builder utilises Hugo’s Markdown extension for highlighting code syntax. The code theme can be selected in the config/_default/params.yaml file.

```python
import pandas as pd
data = pd.read_csv("data.csv")
data.head()
```

renders as

import pandas as pd
data = pd.read_csv("data.csv")
data.head()

Inline Images

{{< icon name="python" >}} Python

renders as

Python

Learn Python

Tue, 24 Oct 2023 00:00:00 +0000

Hugo Blox Builder is designed to give technical content creators a seamless experience. You can focus on the content and the Hugo Blox Builder which this template is built upon handles the rest.

Embed videos, podcasts, code, LaTeX math, and even test students!

On this page, you’ll find some examples of the types of technical content that can be rendered with Hugo Blox.

Video

Teach your course by sharing videos with your students. Choose from one of the following approaches:

Youtube:

{{< youtube w7Ft2ymGmfc >}}

Bilibili:

{{< bilibili id="BV1WV4y1r7DF" >}}

Video file

Videos may be added to a page by either placing them in your assets/media/ media library or in your page’s folder, and then embedding them with the video shortcode:

{{< video src="my_video.mp4" controls="yes" >}}

Podcast

You can add a podcast or music to a page by placing the MP3 file in the page’s folder or the media library folder and then embedding the audio on your page with the audio shortcode:

{{< audio src="ambient-piano.mp3" >}}

Try it out:

Test students

Provide a simple yet fun self-assessment by revealing the solutions to challenges with the spoiler shortcode:

{{< spoiler text="👉 Click to view the solution" >}}
You found me!
{{< /spoiler >}}

renders as

👉 Click to view the solution

You found me 🎉

Math

Hugo Blox Builder supports a Markdown extension for $\LaTeX$ math. You can enable this feature by toggling the math option in your config/_default/params.yaml file.

To render inline or block math, wrap your LaTeX math with {{< math >}}$...${{< /math >}} or {{< math >}}$$...$${{< /math >}}, respectively.

We wrap the LaTeX math in the Hugo Blox math shortcode to prevent Hugo rendering our math as Markdown.

Example math block:

{{< math >}}
$$
\gamma_{n} = \frac{ \left | \left (\mathbf x_{n} - \mathbf x_{n-1} \right )^T \left [\nabla F (\mathbf x_{n}) - \nabla F (\mathbf x_{n-1}) \right ] \right |}{\left \|\nabla F(\mathbf{x}_{n}) - \nabla F(\mathbf{x}_{n-1}) \right \|^2}
$$
{{< /math >}}

renders as

Example inline math {{< math >}}$\nabla F(\mathbf{x}_{n})${{< /math >}} renders as $\nabla F(\mathbf{x}_{n})$ .

Example multi-line math using the math linebreak (\\):

{{< math >}}
$$f(k;p_{0}^{*}) = \begin{cases}p_{0}^{*} & \text{if }k=1, \\
1-p_{0}^{*} & \text{if }k=0.\end{cases}$$
{{< /math >}}

renders as

$$ f(k;p_{0}^{*}) = \begin{cases}p_{0}^{*} & \text{if }k=1, \\ 1-p_{0}^{*} & \text{if }k=0.\end{cases} $$

Code

Hugo Blox Builder utilises Hugo’s Markdown extension for highlighting code syntax. The code theme can be selected in the config/_default/params.yaml file.

```python
import pandas as pd
data = pd.read_csv("data.csv")
data.head()
```

renders as

import pandas as pd
data = pd.read_csv("data.csv")
data.head()

Inline Images

{{< icon name="python" >}} Python

renders as

Python

Paper List

Tue, 24 Oct 2023 00:00:00 +0000

Yufei Liu's Homepages

Example Talk

FLUX.2-based Image Identity Consistency Enhancement

Qualitative Results

Frontal Face Reshaping × JJ Lin

Half-Profile Face Reshaping × Jackson Yee

Strong Stage Lighting × G-Dragon

Overall Evaluation

SynPo: Boosting Training-Free Few-Shot Medical Segmentation via High-Quality Negative Prompts

OmniPersona: Holistic Identity-Preserving Video Creation and Editing via Spatiotemporal Decoupled Persona Injection in Unified DiT

Overview

Generation Example

OneFont: A Unified Agent for End-to-End Font Creation

From Manual Model Selection to OneFont

Motivation

Method Overview

Qualitative Results

Wan-VACE-based Video Object and Shadow Removal

Condition Video vs. Baseline

Condition Video vs. Ours

Clinical Study on the Application of a High-Sensitivity Electronic Nose on Thin-Film Gas Sensor Array Technology Combined with Deep Learning Algorithm for Early Non-Invasive Diagnosis of Chronic Atrophic Gastritis

TEmory: A Temporal-Memory Approach to Weakly Supervised Colonic Polyp Frame Detection

Projects

An Interpretable Approach to the Solutions of High-Dimensional Partial Differential Equations

Learn JavaScript

Video

Podcast

Test students

Math

Code

Inline Images

Did you find this page helpful? Consider sharing it 🙌

Learn Python

Video

Podcast

Test students

Math

Code

Inline Images

Did you find this page helpful? Consider sharing it 🙌

Paper List