deepinsight /insightface
State-of-the-art 2D and 3D Face Analysis Project
★ 30,228⑂ 6,134👁 518⑃ 226● 2,627▤ 2
+3
Daily
+98
Weekly
+248
Monthly
30,228
stars
6,134
forks
518
watchers
226
prs
2,627
issues
2
releases
README
InsightFace: 2D and 3D Face Analysis Project InsightFace project is mainly maintained by Jia Guo and Jiankang Deng. For more information, please visit our website at https://www.insightface.ai InsightFace 2.0 Update InsightFace 2.0 adds PrivateFrame for local video face blur and mosaic, reference-photo selection, and editable analysis JSON through desktop, CLI, and Python workflows, plus optional RGB liveness detection for FaceAnalysis and InsightFace Server. The Python package also supports raccoon_s / raccoon_l model packages, automatic CoreML/CUDA/CPU provider selection, and reusable CoreML compilation caches. PrivateFrame defaults to Fast mode (target 15 analysis FPS). PersonAnalysis adds person detection, face recognition, and body ReID for matching people against registered references. Use its Python API or try local videos, webcams, and RTSP streams in the desktop GUI. For installation and use, start with the Python SDK and GUI documentation. InsightFace Server The new InsightFace Server provides a simple self-hosted Web UI, snake_case REST API, and Python client for detection, optional liveness checks, comparison, registration, and Person search. A single Linux x86_64 CPU or NVIDIA GPU container runs local ONNX Runtime inference and SQLite with operator-supplied, manifest-verified models. It is a simple AWS Rekognition alternative with accuracy-preserving INT8 embedding quantization and 50M+ image search on one RTX 5090 GPU. License The code of InsightFace is released under the MIT License. There is no limitation for both academic and commercial usage. The training data containing the annotation (and the models trained with these data) are available for non-commercial research purposes only. Both manual-downloading models from our github repo and auto-downloading models with our python-library follow the above license policy(which is for non-commercial research purposes only). 2025-11-24 Update: 1. For inswapper series face swap models (e.g., inswapper_128.onnx/inswapper-512-live), please contact [email protected] for licensing and additional support. 2. For open-sourced face recognition models (e.g., buffalo_l package), please contact [email protected] for licensing. 3. For advanced face recognition SDK and models (e.g., InspireFace SDK), please contact [email protected] for licensing and additional support. ChangeLogs 2026-10-03 InsightFace 2.1 PersonAnalysis update: Add person detection, face recognition and body ReID with separate get(), match() and optional update() calls. Support cheetah_s / cheetah_l on CPU or CUDA with automatic model downloads, plus a GUI workflow for local videos, webcams and RTSP with analysis-rate controls and advanced settings. See the Python package and full change log. 2026-09-09 InsightFace 2.0 Liveness update: Add the optional RGB liveness addon, normal/observe modes, per-face scores and input-rejection reasons, plus Server Web model installation an
Top issues
#1564[MFR21] ICCV21 masked face recognition challenge discussion105 comments · 2021-06-04#2469Licensing and how to NOT violate?43 comments · 2023-11-14#1525AssertionError:assert 'detection' in self.models 34 comments · 2021-05-18#2430ERROR: Could not build wheels for insightface, which is required to install pyproject.toml-based projects31 comments · 2023-09-12#1518SCRFD issues29 comments · 2021-05-15#2294Cannot download model "'inswapper_128.onnx'"27 comments · 2023-05-09#1069Instant discussion QQ/WeChat group27 comments · 2020-02-21#2043Unable to Replicate ViT+WebFace42M model results21 comments · 2022-07-04
Latest release
InsightFace Model Zoo model-zoo 2026-09-03This release provides official InsightFace model packs and pretrained models. Model versions are managed independently. Updates to one model do not imply updates to other models. Existing legacy model files may also be mirrored here for centralized distribution. Legacy model files mirrored here are identical to their previously released versions unless explicitly stated otherwise.