Skip to content

Latest commit

 

History

History
223 lines (218 loc) · 11.8 KB

File metadata and controls

223 lines (218 loc) · 11.8 KB
layout fancy_home
permalink /publications-visual/
title Research Atlas

Curated Research Map

Systems, Agents, and Low-Level Vision

Three streams drive my current research. AIGC systems push structured creativity, vision-language agents reason over complex scenes, and low-level restoration builds the reliable backbone beneath both. The gallery below highlights representative works in each space.

AIGC Vision-Language Agents Low-Level Restoration
6
Top-tier venues
ICLR / CVPR / ICCV / AAAI
3
Research axes
AIGC · VLM · Low-Level
Creativity
Task-driven design

AIGC Systems

Structure-Aware Poster Generation

Unified workflows for controllable poster generation, combining task distillation, reward modeling, and data curation.

Flagship Venues ICLR'26 · arXiv'26
PosterOmni
Poster Design · arXiv 2026

PosterOmni — Generalized Artistic Poster Creation via Task Distillation and Unified Reward Feedback

A generalist system that distills local editing, global layout, and reward feedback into a single controllable workflow with public demos and datasets.

PosterCraft
ICLR 2026 · Unified Diffusion

PosterCraft — Rethinking High-Quality Aesthetic Poster Generation in a Unified Framework

Accepted by ICLR'26, PosterCraft couples layout planning and stylized diffusion to translate prompts directly into production-ready posters with consistent typography.

GPT-4o
Evaluation · arXiv 2025

An Empirical Study of GPT-4o Image Generation Capabilities

A 360° evaluation of GPT-4o image generation, benchmarking fidelity, controllability, and safety to guide industrial adoption.

Vision-Language & Agents

Reasoning-Centric Restoration Agents

MLLMs and autonomous agents coordinate perception, planning, and feedback loops for adverse-weather driving and iterative data engines.

Flagship Venues CVPR'25
SnowMaster
CVPR 2025 · Multi-Model Feedback

SnowMaster — Comprehensive Real-world Image Desnowing via MLLM with Multi-Model Feedback Optimization

Combines reasoning from multiple language-vision experts with visual priors to schedule desnowing operations that adapt to scene structure.

JarvisIR
CVPR 2025 · Autonomous Driving

JarvisIR — Elevating Autonomous Driving Perception with Intelligent Image Restoration

An intelligent restoration agent that reasons via MLLM dialogs, calling specialized tools to deweather automotive perception stacks.

Detect Any Mirrors
CVPR 2025 · Data Engines

Detect Any Mirrors — Boosting Learning Reliability with an Iterative Data Engine

Uses agentic feedback and large-scale pseudo-labeling to tackle mirror detection, providing a blueprint for general VLM data refinement.

Low-Level Vision

Physics-aware Restoration Pipelines

Pairing generative priors with physical constraints produces reliable de-weathering, low-light enhancement, and deraining solutions.

Flagship Venues ICCV'25 · AAAI'25
GenHaze
ICCV 2025 · Controllable Haze

GenHaze — One-step Controllable Haze Generation for Real-World Dehazing

Provides a generative counterpart to dehazing by synthesizing paired data with explicit physical controls, enabling better restoration agents.

PromptHaze
AAAI 2025 · Prompted Dehazing

PromptHaze — Prompting Real-world Dehazing via Depth Anything

Aligns promptable depth priors with restoration networks, achieving plug-and-play robustness across weather distributions.

AGLLDiff
arXiv 2024 · Low-Light Diffusion

AGLLDiff — Guiding Diffusion Models for Training-Free Low-Light Enhancement

Introduces adaptive guidance so diffusion models can enhance low-light scenes without paired training data.

<style> .pub-hero { display: flex; flex-wrap: wrap; gap: 1.5rem; background: linear-gradient(120deg, #0f172a, #1f2937); color: #f8fafc; padding: 2.5rem; border-radius: 24px; box-shadow: 0 20px 60px rgba(15, 23, 42, 0.45); } .pub-hero .hero-text {flex: 1 1 320px;} .hero-text h1 {margin: 0.4rem 0 0.8rem; font-size: 2.3rem;} .hero-tags span { display: inline-block; margin-right: 0.5rem; margin-bottom: 0.5rem; padding: 0.35rem 0.9rem; border-radius: 999px; background: rgba(255,255,255,0.15); font-size: 0.85rem; } .pub-hero .hero-notes {display: flex; flex: 1 1 240px; gap: 1rem; align-items: stretch;} .note-card {flex: 1; background: rgba(255,255,255,0.08); border-radius: 18px; padding: 1.2rem; text-align: center;} .note-value {font-size: 2.6rem; font-weight: 700; line-height: 1;} .note-label {font-size: 0.95rem; opacity: 0.85; margin-top: 0.4rem;} .eyebrow {text-transform: uppercase; letter-spacing: 0.08em; font-size: 0.9rem; color: #94a3b8; margin: 0;} .category-grid {margin-top: 2.5rem; display: flex; flex-direction: column; gap: 2.5rem;} .category-section {background: #f8fafc; border: 1px solid #e2e8f0; border-radius: 24px; padding: 2rem;} .category-headline {display: flex; flex-wrap: wrap; gap: 1.5rem; justify-content: space-between; align-items: center;} .category-headline h2 {margin: 0.4rem 0; font-size: 1.8rem;} .category-stat {text-align: right; min-width: 180px;} .stat-label {display: block; font-size: 0.85rem; color: #475569; text-transform: uppercase; letter-spacing: 0.08em;} .stat-value {font-size: 1.1rem; font-weight: 600; color: #0f172a;} .paper-grid {margin-top: 1.5rem; display: grid; grid-template-columns: repeat(auto-fit, minmax(260px, 1fr)); gap: 1.5rem;} .paper-card {background: #ffffff; border-radius: 20px; padding: 1.2rem; display: flex; flex-direction: column; box-shadow: 0 12px 35px rgba(15, 23, 42, 0.08); border: 1px solid #e2e8f0;} .paper-image {width: 100%; height: 160px; object-fit: cover; border-radius: 16px; margin-bottom: 1rem;} .paper-meta {font-size: 0.85rem; text-transform: uppercase; letter-spacing: 0.08em; color: #64748b;} .paper-card h3 {font-size: 1.15rem; margin: 0.4rem 0 0.6rem; color: #0f172a;} .paper-card p {flex: 1; color: #475569; font-size: 0.95rem;} .paper-links {margin-top: 1rem; display: flex; flex-wrap: wrap; gap: 0.5rem;} .pill-link {padding: 0.35rem 0.9rem; border-radius: 999px; background: #0f172a; color: #fff; font-size: 0.85rem; text-decoration: none; transition: transform 0.2s ease, background 0.2s ease;} .pill-link:hover {transform: translateY(-2px); background: #1d4ed8;} @media (max-width: 768px) { .pub-hero {flex-direction: column;} .hero-notes {width: 100%;} .category-headline {flex-direction: column; align-items: flex-start;} .category-stat {text-align: left;} } </style>