# Third-party notices This file tracks production dependencies and model artifacts for the release candidate. Formal commercial redistribution review remains required where explicitly noted below. ## Python runtime dependencies Exact versions, license metadata, and the license files shipped by each wheel are bundled in `third_party_licenses/`. The release build collects these files for NumPy, ONNX Runtime, OpenCV, Pillow, piexif, platformdirs, pyclipper, pywebview, and Shapely. NumPy's bundled-library notices are preserved in its collected license file. ## Microsoft Edge WebView2 Evergreen Bootstrapper - Publisher: Microsoft Corporation - Artifact: `MicrosoftEdgeWebview2Setup.exe` - Usage: installed only when the Evergreen Runtime is absent - Distribution documentation: - The bundled executable must retain a valid Microsoft Authenticode signature. ## Inno Setup Inno Setup 7.0.2 is used as a build tool and is not included as an application runtime component. The resulting installer uses its official Simplified Chinese translation. Commercial distribution requires an applicable Inno Setup commercial license; this remains an external release gate. ## Lucide icons - Project: - Usage: a local, release-bundled subset of interface icons under `app/web/assets/icons/` - License: ISC - Copyright: Lucide Contributors - Runtime network access: none ## PaddleX PP-YOLOE-S vehicle detector - Project: - Runtime artifact: `ppyoloe_vehicle.onnx` - Derived from: `PP-YOLOE-S_vehicle_infer.tar` - Source: - SHA-256: `1143b3e62e1716ed056870f3788da77457b932cc7efa10b3abca5d24f61d0b2e` - Project license: Apache License 2.0 - Model description: 640 x 640 single-class PP-YOLOE-S vehicle detector. - Conversion: Paddle2ONNX 2.1.0, ONNX opset 14, no retraining. - Release note: the downloaded model archive contains no separate license file. Redistribution must receive a final terms review before packaging. ## PaddleDetection PP-Vehicle plate detector - Project: - Pinned documentation revision: `b25522a0f4bde8c80603f3ba5e3472059972e3b5` - Runtime artifact: `ppocrv3_plate.onnx` - Derived from: `ch_PP-OCRv3_det_infer.tar.gz` - Source: - SHA-256: `acc7eb42b299cdb4eed2999f4de99c89555767b321c272a9878f688d24503fd9` - Project license: Apache License 2.0 - Model description: PP-OCRv3 text detector fine-tuned on a mixture of CCPD2019 and CCPD2020 for plate detection. - Conversion: Paddle2ONNX 2.1.0, ONNX opset 14, no retraining. - Release note: the downloaded model archive contains no separate license file. Redistribution must receive a final terms review before packaging. ## MI-GAN inpainting model - Project: - Artifact: `migan_pipeline_v2.onnx` - Source: - SHA-256: `6f1f3530a1a2324b19752018ce756088b07973cda8d7d890034ace5c8a48c40b` - Project license: MIT - Model description: MI-GAN 512 Places2 ONNX pipeline linked by the official repository and hosted by a paper author. - Product status: rejected by visual quality review and disabled. - Release note: redistribution must still receive a final terms review before packaging. ## OpenCV quantized LaMa inpainting model - Project: - Artifact: `inpainting_lama_2025jan.onnx` - SHA-256: `7df918ac3921d3daf0aae1d219776cf0dc4e4935f035af81841b40adcf74fdf2` - Publisher: Open Source Vision Foundation OpenCV account - License: Apache License 2.0 - Product status: selected reference inpainter after visual comparison. ## PaddleOCR DB post-processing - Project: - Reference revision: `2661c7c0ef5c613e8f93c6e93b2e052399f0f854` - Referenced files: `ppocr/data/imaug/operators.py`, `ppocr/postprocess/db_postprocess.py` - License: Apache License 2.0 - Usage: the project contains an independently maintained, reduced adaptation of the DB resize, normalization, contour scoring, and unclip algorithms.