Scalp Checker

Check your diagnosis with XAI (Grad-CAM for CNN, Attention HeatMap for ViT) to understand model's focus.

Ha, Changjin, Taesik Go, and Woorak Choi. "Intelligent Healthcare Platform for Diagnosis of Scalp and Hair Disorders." Applied Sciences 14.5 (2024): 1734.

🖥️ How to Use

• Download&Build Project
• Download Project : Scalp Checker Python Project
• Download Project : Scalp Checker GPU Compatibility
• Build Python Project as exe (For me, I've been used PyInstaller)
• Move exe, dll and another build files into
  • Python Project : C:\Program Files\ScalpChecker\include
  • GPU Compatibility : C:\Program Files\ScalpChecker\GPUCompatibility
  • h5 Models : C:\Program Files\ScalpChecker\Models

🚀 Tech Stack

Client (Windows)

• Framework: WPF (.NET 8)
• Architecture: Feature-based modular Structure (Inspection, Dashboard, Frameworks, Statistics)
• Visualization: LiveChartsCore (Real-time scalp health analytics)

AI & Inference

• Models: Keras CNN (Spatial Feature Extraction) & Vision Transformer (ViT - Global Context)
• Target Conditions: 6-class diagnosis (Dandruff, Sebum, Erythema, Fine Hair Loss, Pustule, Alopecia)
• Execution: Subprocess-based inference bridge via external main.exe

🏗️ Architecture

graph TD
    %% Windows Frontend Modules
    subgraph WPFApp [🖥️ ScalpChecker: WPF .NET 8]
        UI[Main Window / UI]
        ModuleCheck[Check Module: Image Handler]
        ModuleDash[Dashboard: LiveChartsCore]
        ModuleStat[Statistics: Session History]
        Framework[Frameworks: Core Logic]

        UI <--> ModuleCheck
        UI <--> ModuleDash
        UI <--> ModuleStat
        ModuleCheck <--> Framework
    end

    %% External Inference Bridge
    subgraph AIEngine [🧠 External Inference Engine]
        MainExe[main.exe / Python Subprocess]
        KerasCNN[CNN Model]
        ViT[Vision Transformer]

        Framework -->|1. Request Inference| MainExe
        MainExe -->|2. Multi-Model Analysis| KerasCNN & ViT
        KerasCNN & ViT -->|3. JSON Result Output| MainExe
        MainExe -->|4. Return Results| Framework
    end

    %% Data Flow
    Framework -->|Update| ModuleDash
    Framework -->|Save| ModuleStat

Loading

🧱 If I were to rebuild it in 2026

Layer	Original	2026 Pick	Reason
UI Framework	WPF + MaterialDesignThemes	WPF + CommunityToolkit.Mvvm or WinUI 3	MVVM Toolkit adds source-generated ViewModels at zero cost; WinUI3 provides the modern components used in the latest Windows on its own.
Inference	Python subprocess + .h5 Keras	ONNX Runtime + DirectML	In-process, no Python dependency, GPU-accelerated on any DirectX 12 device
Charts	LiveChartsCore beta	LiveChartsCore 2.x sable or ScottPlot 5	Drop the beta tag; ScottPlot 5 is lighter
Config / IPC	Flat .txt files	System.Text.Json + appsettings.json	Schema, versioning, one-liner read/write
Packaging	Unpackaged EXE	MSIX via dotnet publish	Bundles models, handles install path correctly, enables Store distribution
Icons	FontAwesome.Sharp	Fluent UI icons (Segoe Fluent Icons)	Ships with Windows 11, zero dependency, correct visual style
Deployment	Manual file copy to Program Files	MSIX Packaging / AppInstaller	Automatic installation and clean updates via Windows standard deployment

✨ Core Features

Show Contents

Dashboard

Statistics of your inspection results.

Inspection

Inspect your Scalp using our EfficientNet or ViT Model, and you can select your inspection parts

Check your Result with Grad-CAM (EfficientNet) or Attention HeatMap(ViT)

History

Review your Inspection Result, Control with date.

Build&Test Environment

Microsoft Windows 11 Pro
Intel Core i7-12700K
32GB RAM
Python 3.9.13
Microsoft Visual Studio 2022
PyCharm 2022.3.1
Tensorflow 2.10.1

System Requirements

CPU : Intel Core i5-7500 or up
RAM : 4GB or up
OS : Microsoft Windows 8.1 (x64) or up
Disk Space : 6GB or up
GPU : N/A (NVIDIA GPU Recommend.)
Others : .NET 8 Runtime Required.
if GPU Available, Please install CUDA + cuDNN + NVIDIA Geforce Driver for your GPU