Elements

PRISM

PCB Recognition and Intelligent Schematic Mapping

PRISM reconstructs an editable KiCad schematic from aligned images of the top and bottom sides of a PCB. It combines image processing, pad and trace detection, OCR-assisted component recognition, connectivity analysis, manual correction tools, and schematic optimization in one desktop application.

PRISM is currently alpha software. Generated connectivity and component assignments must be reviewed before they are used for manufacturing, servicing, or safety-critical work.

Features

• guided import of TOP and BOTTOM PCB images;
• automatic or manually sampled solder-mask and copper colors;
• board alignment, pad pairing, trace detection, and net reconstruction;
• OCR-assisted labels and component candidates;
• interactive correction editor for pads, traces, pairs, OCR, and components;
• text netlist and KiCad schematic export;
• optional KiCad CLI validation and schematic optimization;
• native desktop builds for Windows, macOS, and Linux.

Requirements

• Python 3.11 or newer when running from source;
• Tk 8.6 or newer;
• Tesseract OCR for OCR features;
• KiCad with kicad-cli for round-trip validation and optimized previews.

PRISM searches PATH first and then conventional installation locations for the current operating system. Tesseract can also be selected manually in the interface.

On Debian or Ubuntu, install the native GUI and OpenCV runtime libraries before running from source or building:

sudo apt-get install python3-venv python3-tk libgl1 libglib2.0-0

Run From Source

Create and activate a virtual environment, then install the project:

python -m venv .venv

Windows PowerShell:

.\.venv\Scripts\Activate.ps1
python -m pip install -r requirements.txt
prism

macOS and Linux:

source .venv/bin/activate
python -m pip install -r requirements.txt
prism

Windows users can alternatively run .\run.ps1.

Build Native Packages

PyInstaller creates native applications and does not cross-compile. Run the matching script on each target operating system:

# Windows
.\scripts\build\build_windows.ps1

# macOS
bash scripts/build/build_macos.sh

# Linux
bash scripts/build/build_linux.sh

The scripts create an isolated .venv-build environment and use the shared configuration in scripts/build/package.py.

Windows and Linux use PyInstaller's onedir layout. Distribute the complete dist/PRISM directory, not only the executable. The macOS application is unsigned; public releases should be code-signed and notarized by the release owner.

Basic Workflow

1. Select images of the top and bottom PCB sides.
2. Confirm automatic color detection or collect color samples manually.
3. Run the analysis.
4. Review detected geometry and connectivity in the correction editor.
5. Recalculate after corrections and inspect the generated netlist.
6. Open the result directory and validate the KiCad schematic.

Generated files are written under ~/PRISM/results by default. Set the PRISM_OUTPUT_DIR environment variable to use a different location. A local results/ directory is also excluded from version control.

Dataset

Dataset/ contains twelve PCB examples with TOP/BOTTOM images and reference KiCad projects. It is intended for repeatable development and evaluation, not as a formal train/test split.

Reproducible Evaluation

evaluation/ contains the versioned reference annotations and English-language tools for synthetic regression tests and ablation studies. A baseline run is started from the repository root with:

python evaluation/run_evaluation.py

The six supported ablation configurations can be executed with:

python evaluation/run_ablations.py

The tools generate machine-readable JSON reports, compact Markdown summaries, and a LaTeX ablation table under evaluation/results/. See the evaluation guide for sample selection, dependency requirements, command-line options, report interpretation, and the exact scope of the synthetic results. These tests exercise controlled KiCad-derived board images; they do not validate PRISM on photographs of physical PCBs.

Project Structure

.
|-- Dataset/                         Test images and reference KiCad projects
|-- evaluation/                      Synthetic evaluation, ablations, and reference annotations
|-- scripts/build/                   Native build entry points and shared build logic
|-- src/schematic_generator/
|   |-- assets/                      Application and executable icons
|   |-- correction_editor/           Manual reconstruction editor
|   |-- detection/                   Pad, trace, plane, and pair detection
|   |-- diagnostics/                 Diagnostic reports
|   |-- geometry/                    Alignment and rectification
|   |-- gui/                         Main desktop workflow
|   |-- kicad/                       KiCad serialization and routing
|   |-- netlist/                     Connectivity and component inference
|   |-- schematic/                   Validation and optimization
|   |-- platform_support.py          Operating-system integration
|   `-- __main__.py                  Application entry point
|-- pyproject.toml                  Package metadata and runtime dependencies
|-- requirements-build.txt          Build-only dependencies
`-- requirements.txt                Editable development installation

Automated Builds

The GitHub Actions workflow in .github/workflows/build.yml runs smoke tests and creates Windows, macOS, and Linux artifacts using the same scripts as local builds.

Contributing

Keep platform-independent behavior in the application modules. Operating-system differences belong in platform_support.py, while packaging differences belong in scripts/build/package.py. New detection or reconstruction behavior should be checked against multiple samples from Dataset/.

License

No project license has been selected yet. Add a LICENSE file before presenting PRISM as open-source software or accepting external contributions.