Simplevision

Build OpenCV pipelines graphically. Run them anywhere.

SimpleVision is a free, open-source educational tool for students learning computer vision. It lets you assemble image-processing pipelines step by step, preview each stage on a real frame, and export the result as portable JSON or runnable Python. Pipelines run locally through a bundled Python interpreter with OpenCV and numpy — no installation, no cloud, no network.

Windows installers ship today; macOS and Linux are on the roadmap. Auto-updates land via GitHub Releases.

Features

Visual Pipeline Editor

• Drag-and-drop steps from a categorized function library — 22 operations covering acquisition, color, filtering, thresholding, morphology, and analysis
• Live frame-by-frame preview — every step renders its own intermediate result so you can see exactly what the pipeline does
• Per-step inspector with paper-tab Setup panels and contextual parameter controls
• SVG overlays for detection results (bounding boxes, particle markers) layered on top of the source frame, not baked in

Export Anywhere

• .simplevision — portable, schema-versioned pipeline definitions (JSON inside) that load in any SimpleVision install
• .py codegen — drop-in Python scripts using cv2 calls, ready to run outside the app
• Pipelines are deterministic: same JSON + same input = same output, on any machine

Self-Contained Runtime

• Bundled Python interpreter (via python-build-standalone) with OpenCV and numpy pre-installed
• End users don't need to install Python, pip, or anything else — the installer brings everything
• Sidecar process spawned per pipeline run; results streamed back over stdout

Desktop Polish

• Light-mode UI inspired by industrial vision tools, built with custom design tokens and the Geist typeface
• Auto-updater wired up against GitHub Releases — users get new versions on next launch
• Native file dialogs, no flashing console windows, GUI-subsystem release builds

Plugins

SimpleVision will soon support pre-trained AI models — segmentation, depth, detection — directly inside its pipeline editor. Open the Plugins menu to see what's available. (The first plugins will land in a follow-up release.)

Function Library

Architecture

   ┌─────────────────────────────────────────────────────┐
   │                  Tauri Desktop Shell                │
   │                                                     │
   │   ┌─────────────────────────────────────────────┐   │
   │   │            Svelte 5 Frontend                │   │
   │   │                                             │   │
   │   │   Library  │  Pipeline  │  Viewer  │ Setup  │   │
   │   │                                             │   │
   │   │   Cubits (state) · Repositories · DI        │   │
   │   └────────────────────┬────────────────────────┘   │
   │                        │  Tauri IPC                 │
   │   ┌────────────────────┴────────────────────────┐   │
   │   │              Rust Core (src-tauri)          │   │
   │   │   File I/O · Sidecar lifecycle · Updater    │   │
   │   └────────────────────┬────────────────────────┘   │
   └────────────────────────┼────────────────────────────┘
                            │  stdin / stdout
                            │  (JSON pipeline → PNG + JSON results)
                   ┌────────┴────────┐
                   │  Python Sidecar │
                   │                 │
                   │  cv2  ·  numpy  │
                   │  simplevision_  │
                   │  runtime        │
                   └─────────────────┘

Workspace Layout

Data Flow

 .simplevision  ──spawn──▸  python sidecar  ──load──▸  cv2 ops  ──write──▸  step_N.png
   (pipeline)            (stdin JSON)               (per step)             results.json
                                                                              │
                                                                              ▼
                                                                       SVG overlays
                                                                       in viewer

The frontend never calls OpenCV directly. Every pipeline run spawns a fresh Python process, feeds it the JSON, and reads back per-step output files from a sandboxed working directory next to the input frame.

Getting Started

Prerequisites

• Node 20+ with pnpm 11+ (via corepack enable pnpm)
• Rust stable + cargo (curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh)
• Python 3.11+ (for the dev venv — release builds bundle their own)

Quick Start

pnpm install
pnpm setup:runtime      # builds runtime/.venv with OpenCV + numpy
pnpm tauri dev

The editor opens with the function library on the left and an empty pipeline ready to assemble.

[!TIP] Drop any .png or .jpg on the viewer to load it as the source frame. Add a Load Image step from the library to make the source frame part of the pipeline definition.

bash scripts/build-sidecar.sh   # one-time, populates app/src-tauri/python-runtime/
pnpm tauri build                # produces .msi + .nsis on Windows

SimpleVision/
├── app/
│   ├── src/
│   │   └── lib/
│   │       ├── components/      # Svelte UI (library, pipeline, viewer, setup)
│   │       ├── cubits/          # State containers (pipeline, viewer, ...)
│   │       ├── repositories/    # File I/O + pipeline persistence
│   │       ├── pipeline/        # Pipeline execution coordination
│   │       ├── export/          # JSON + Python codegen
│   │       ├── updater/         # Auto-update flow
│   │       └── data/            # Function catalogue
│   └── src-tauri/               # Rust shell (Tauri 2) + bundled python-runtime
├── runtime/
│   └── simplevision/            # Python package (student API + execution engine)
├── scripts/
│   ├── setup-runtime.sh         # Local dev venv
│   └── build-sidecar.sh         # Portable python-build-standalone bundle
├── samples/                     # Reference images + example pipelines
├── docs/                        # Pipeline spec, design plans
└── .github/workflows/           # release.yml, cache-warm.yml

CI / CD

Roadmap

• macOS and Linux release builds (Windows shipping today)
• Pattern matching op (template-based, with a "Create Template" flow)
• Arithmetic ops — chain B − R style channel math
• Histogram visualization in the inspector
• Multi-pipeline projects with tabs
• Native menus + accelerators (Ctrl+S, Ctrl+O)

License

Apache License 2.0 — see LICENSE.

SimpleVision is an independent, open-source educational project. It is not affiliated with, endorsed by, or derived from any commercial computer vision tool. All trademarks are the property of their respective owners.