Skip to content
View wanxiayushaonian's full-sized avatar

Block or report wanxiayushaonian

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
wanxiayushaonian/README.md

Typing SVG


GitHub Stats

Top Languages

About Me

class DeepLearningResearcher:
    def __init__(self):
        self.name = "ljb"
        self.role = "Deep Learning Researcher & AI Agent Builder"
        self.research = [
            "HDR Image Reconstruction",
            "Intrinsic Image Decomposition",
            "Diffusion Models",
            "Flow Matching",
        ]
        self.tools = [
            "Claude Code", "OpenAI Codex",
            "Custom Skills", "MCP Servers",
        ]

    def daily_routine(self):
        return [
            "📖 Reading papers from CVPR / ICCV / ECCV / ICLR",
            "🧪 Training models & running experiments",
            "🤖 Building AI-powered development tools",
            "💻 Shipping full-stack applications",
        ]

Tech Stack

Deep Learning & Computer Vision

PyTorch HuggingFace OpenCV NumPy Pandas

LLM & Agent Development

Anthropic LangChain

Languages & Frameworks

Python Kotlin TypeScript C++ Java

Backend & DevOps

FastAPI Docker Git Linux Hydra

Mobile & Frontend

Android Jetpack Compose Vite

AI Coding Tools

Claude Code Cursor

Featured Projects

MEF-IID

Self-supervised HDR reconstruction via intrinsic image decomposition

  • Retinex physics model: I = A x S + R
  • InstanceNorm dual-path encoder
  • RAFT optical flow in albedo domain
  • 3-phase curriculum training, 11 loss terms
  • ~12M params, 50ms/frame (20 FPS)

PyTorch EfficientNet-B3 RAFT Hydra

Nexus

AI cognitive companion with 7 specialized agents

  • Full-stack: FastAPI + TypeScript + SQLite
  • 7 AI agents (Distillation, Socratic, Debate...)
  • L2 canvas tools + L3 agent tools architecture
  • SSE streaming, multi-round tool calling
  • Docker deployment, token auth

FastAPI Anthropic SDK TypeScript Docker

Diffusion Model Research

Reproducing SOTA diffusion-based imaging methods

  • UltraFusion (CVPR 2025): SD v2.1 + ControlNet
  • PhysHDR: SD v1.5 + physics-aware conditioning
  • Flow Matching: 7 sub-projects, CNF + ODE
  • PnP-Flow (ICLR 2025) reproduction

PyTorch Diffusers ControlNet CLIP

MindCard

On-device AI inspiration management (Android)

  • Kotlin + Jetpack Compose + Room
  • Gemma 3 1B Q4 on-device inference via MediaPipe
  • Auto title generation, keyword extraction
  • "Information distillation" interaction pattern
  • Fully offline, ~554MB model

Kotlin MediaPipe Gemma MVVM

Research Focus

HDR Image Reconstruction ─┬─ Multi-Exposure Fusion (MEF)
                          ├─ Intrinsic Image Decomposition (IID)
                          ├─ Diffusion Models (LDM, ControlNet)
                          └─ Flow Matching

AI Agents ────────────────┬─ LLM Tool Calling & Multi-Agent Systems
                          ├─ On-device LLM Inference
                          └─ AI-Powered Development Tools

GitHub Activity


GitHub Streak

Let's Connect

Email GitHub


Profile Views

Popular repositories Loading

  1. LFM-HDR-DiT LFM-HDR-DiT Public

  2. IIDMEF IIDMEF Public

    Python

  3. softwareTest softwareTest Public

    my test code for jenkins continue test

    Java

  4. IST IST Public

    Python

  5. MentorOS MentorOS Public

    MentorOS 是一个“全球知识互联的学习与创造伙伴”智能体。它通过认知建模、多风格知识重构、动态知识图谱与脚手架式创造辅助，为每个用户构建一座不断生长的思想宫殿，实现真正的自适应终身学习与深度创造。

    Python

  6. OuterBrainSystem OuterBrainSystem Public

    TypeScript