English | 中文
A WPF desktop application for automated garment pattern extraction and mockup generation, powered by the U-2-Net AI model.
- AI Background Removal — U-2-Net ONNX inference extracts patterns from garment photos with a transparent background.
- Batch Mockup Synthesis — SkiaSharp Multiply-blend compositing overlays extracted patterns onto template garments.
- Quality Control — Side-by-side and overlay comparison canvas with linked zoom/pan; approve or reject each result.
- Pipeline Dashboard — Real-time file counts across the six-folder POD workflow.
- Multilingual UI — Runtime language switching: 简体中文, English, Français, 繁體中文.
Drop garment photos
↓
00_提图队列 (Extract Queue) ← Drop source images here
↓ [AI Extraction – U-2-Net ONNX, removes background]
01_提图完成 (Extract Done) → Transparent-background PNG
↓ (copy/move files to continue)
30_抠图队列 (Refine Queue) ← Manual refinement input
↓ [Optional manual edge-refine step]
31_抠图完成 (Refine Done)
↓
50_成品队列 (Finalize Queue) ← Patterns ready for mockup
↓ [Batch Mockup – SkiaSharp Multiply blend]
51_成品完成 (Finalize Done) → Final composited mockups
↓ [QC Review]
Production_Ready / Rework
All six folders are created automatically when you set the Root Path and click Apply or enable Watch mode. The repo ships with a workspace/ directory containing .gitkeep placeholders so the structure is preserved on clone.
ImageAvatar uses U-2-Net for background removal. The model is not bundled with the repo (≈ 176 MB).
# Download u2net.onnx from the rembg project
https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2net.onnxPlace the file at:
%AppData%\ImageAvatar\models\u2net.onnx
Or configure a custom path in Settings → AI Model → Browse….
| Property | Value |
|---|---|
| Input | [1, 3, 320, 320] float32 |
| Normalize | ImageNet mean/std (RGB) |
| Output | [1, 1, 320, 320] float32 |
| Runtime | Microsoft.ML.OnnxRuntime |
The app reads startup defaults from appsettings.json in the executable directory. This file is ideal for POD environment deployment where paths are fixed across workstations.
{
"ImageAvatar": {
"WorkspaceRoot": "\\\\NAS01\\POD\\workspace",
"ModelPath": "\\\\NAS01\\POD\\models\\u2net.onnx",
"TemplatesFolder": "\\\\NAS01\\POD\\templates"
}
}Leave values empty to use per-user defaults (%Documents%\ImageAvatar). Runtime changes saved through the Settings page are written to %AppData%\ImageAvatar\settings.json and take precedence over appsettings.json.
- Open Settings → Language.
- Select a language from the dropdown:
zh-CN,en-US,fr-FR,zh-HK. - Click Apply — the UI updates immediately without restart.
Prerequisites: .NET 8 SDK, Visual Studio 2022 17.8+
git clone https://github.com/Forrest-tech/ImageAvatar.git
cd ImageAvatar
dotnet restore
dotnet build -c ReleaseRun:
dotnet run --project ImageAvatar/ImageAvatar.csprojThe workspace skeleton is at workspace/ relative to the repo root. Point the Dashboard Root Path to that directory to use it immediately.
| Layer | Library |
|---|---|
| UI Framework | WPF (.NET 8) + WPF-UI 3.x |
| MVVM | CommunityToolkit.Mvvm 8.x |
| DI / Config | Microsoft.Extensions.Hosting |
| AI Inference | Microsoft.ML.OnnxRuntime 1.20 |
| Image Processing | OpenCvSharp4 |
| Canvas / Blend | SkiaSharp 2.88 |
English | 中文
基于 U-2-Net AI 模型的服装图案自动提取与效果图生成桌面应用（WPF）。
- AI 去背景 — 使用 U-2-Net ONNX 模型自动去除服装图片背景，输出透明底 PNG。
- 批量合成效果图 — SkiaSharp Multiply 混合模式将图案叠加到模板服装上，保留布料褶皱与阴影。
- 质检审核 — 支持并排与叠加对比，画布联动缩放/平移，一键通过或退回。
- 流水线仪表盘 — 实时显示六个文件夹的文件数量。
- 多语言界面 — 运行时切换语言：简体中文、English、Français、繁體中文。
投入服装照片
↓
00_提图队列 ← 将原始服装图片放入此处
↓ [AI 提图 – U-2-Net ONNX 去背景]
01_提图完成 → 透明底 PNG
↓
30_抠图队列 ← 需人工精修的图片
↓ [可选：手动精修边缘]
31_抠图完成
↓
50_成品队列 ← 准备合成效果图的图案
↓ [批量合成 – SkiaSharp Multiply 叠加]
51_成品完成 → 最终效果图
↓ [质检审核]
Production_Ready（生产就绪）/ Rework（返工）
设置根目录并点击应用或开启监听模式后，六个文件夹将自动创建。仓库中的 workspace/ 目录包含 .gitkeep 占位文件，克隆后即保留目录结构。
ImageAvatar 使用 U-2-Net 进行去背景处理。模型文件不包含在仓库中（约 176 MB）。
# 从 rembg 项目下载 u2net.onnx
https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2net.onnx将文件放置于：
%AppData%\ImageAvatar\models\u2net.onnx
或在设置 → AI 模型 → 浏览… 中配置自定义路径。
应用启动时从可执行文件目录下的 appsettings.json 读取默认配置，适用于 POD 生产环境统一部署（如网络共享路径）：
{
"ImageAvatar": {
"WorkspaceRoot": "\\\\NAS01\\POD\\workspace",
"ModelPath": "\\\\NAS01\\POD\\models\\u2net.onnx",
"TemplatesFolder": "\\\\NAS01\\POD\\templates"
}
}留空则使用默认路径（%Documents%\ImageAvatar）。通过设置页面保存的配置写入 %AppData%\ImageAvatar\settings.json，优先级高于 appsettings.json。
- 打开设置 → 语言。
- 从下拉框选择语言：
zh-CN、en-US、fr-FR、zh-HK。 - 点击应用 — 界面立即更新，无需重启。
环境要求： .NET 8 SDK、Visual Studio 2022 17.8+
git clone https://github.com/Forrest-tech/ImageAvatar.git
cd ImageAvatar
dotnet restore
dotnet build -c Release仓库根目录的 workspace/ 即为工作流目录，在仪表盘根目录中填入该路径即可立即使用。