{
  "id": "insightface-buffalo-l",
  "identity": {
    "family": "insightface",
    "display_name": "InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K)",
    "hub_id": "insightface/buffalo_l",
    "api_name": null,
    "revision": null,
    "variants": [],
    "license": "InsightFace pretrained models, non-commercial research only (library code MIT)",
    "commercial_use": false,
    "commercial_use_note": "The insightface README and model_zoo state every pretrained model, downloaded automatically or manually, is for non-commercial research only; the MIT license covers the library code. face_identity_extractor bills customers for this pack, so a commercial license from InsightFace is needed.",
    "release_date": null,
    "paper_url": "https://arxiv.org/abs/1801.07698",
    "repo_url": "https://github.com/deepinsight/insightface"
  },
  "task": {
    "category": "face",
    "modalities_in": [
      "image",
      "video"
    ],
    "modalities_out": [
      "boxes",
      "keypoints",
      "vector"
    ],
    "best_at": "Face detection, 5-point alignment and 512-d face identity vectors for finding the same person across images and video frames.",
    "worst_at": "Commercial use (license); tiny or heavily occluded faces below min_face_size; the parameter names do not match the weights: scrfd_2.5g and scrfd_10g both load SCRFD-10GF, and arcface_r100 / magface_r100 load a ResNet50 (w600k_r50), not an R100."
  },
  "io": {
    "input": {
      "image": {
        "det_size": "640x640",
        "align": "5-point affine to 112x112"
      },
      "video": "frames sampled by the extractor, optional dedup of the same face"
    },
    "output": {
      "boxes": "face bounding boxes with detection scores",
      "keypoints": "5 landmarks per face (2d106 and 3d68 landmark models also in the pack)",
      "labels": "gender and age from the pack's genderage head",
      "vector": {
        "dims": 512,
        "distance": "cosine",
        "type": "dense"
      }
    }
  },
  "mixpeek_wiring": {
    "extractors": [
      {
        "extractor": "face_identity_extractor@v1",
        "selected_by": "detection_model=scrfd_2.5g (default) or scrfd_10g; embedding fixed arcface_r100"
      }
    ],
    "feature_uris": [
      "mixpeek://face_identity_extractor@v1/insightface__arcface"
    ],
    "vector_indexes": [
      "face_identity_extractor_v1_embedding"
    ],
    "retriever_stages": [
      "feature_search"
    ],
    "composes_with": [
      "image or video -> SCRFD boxes -> aligned 112x112 crops -> ArcFace 512-d vector per face -> feature_search with a face photo"
    ]
  },
  "run": {
    "status": "served",
    "status_reason": "insightface__arcface Serve app; face_identity_extractor batch calls it (FaceProcessingBatchService, strict_serve, no local fallback).",
    "serve_apps": [
      "insightface__arcface"
    ],
    "runtime": "insightface.app.FaceAnalysis(name=\"buffalo_l\") on onnxruntime, prepare(ctx_id from providers, det_size=(640, 640))",
    "weights_format": "onnx",
    "trust_remote_code": false,
    "size_on_disk_mb": 326,
    "params_millions": null,
    "preprocessing": "SCRFD detection at 640x640, 5-point affine alignment to 112x112.",
    "postprocessing": "ArcFace recognition embedding per aligned face (512 dims), L2-normalized; optional quality scoring and video dedup.",
    "serving_shape": "both",
    "batch_size": 8,
    "serve_batch": null,
    "scale_to_zero": {
      "min_replicas": 0,
      "max_replicas": 4,
      "idle_timeout_s": 600,
      "memory_gb": 2,
      "num_cpus": 0.5,
      "num_gpus": {
        "expr": "hardware_config.gpu_fraction if hardware_config.gpu_concurrency > 0 else 0"
      },
      "cold_start_s": {
        "warm_node_s": null,
        "new_node_s": null,
        "source": null
      },
      "cpu_fallback": true
    }
  },
  "cost": {
    "unmeasured": true,
    "measured": null,
    "note": "No committed measurement. The definition's detection latencies (2-3, 5-7, 10-15 ms) are descriptions, not committed measurements."
  },
  "evidence": {
    "unmeasured": true,
    "items": []
  },
  "refs": [
    "github://deepinsight/insightface/models/buffalo_l.zip@v0.7"
  ],
  "formats": [],
  "use_cases": [
    "face-search"
  ],
  "recommended": {
    "value": false,
    "for": [],
    "reason": null
  },
  "demos": [],
  "seo": {
    "title": "InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K)",
    "description": "InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K) is a face recognition model. Served on Mixpeek, scaling to zero when idle.",
    "keywords": [
      "InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K)",
      "insightface",
      "face recognition",
      "insightface/buffalo_l",
      "face-search"
    ],
    "faqs": [
      {
        "q": "What is InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K) best at?",
        "a": "Face detection, 5-point alignment and 512-d face identity vectors for finding the same person across images and video frames."
      },
      {
        "q": "What are the limits of InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K)?",
        "a": "Commercial use (license); tiny or heavily occluded faces below min_face_size; the parameter names do not match the weights: scrfd_2.5g and scrfd_10g both load SCRFD-10GF, and arcface_r100 / magface_r100 load a ResNet50 (w600k_r50), not an R100."
      },
      {
        "q": "Can InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K) be used commercially?",
        "a": "No. The insightface README and model_zoo state every pretrained model, downloaded automatically or manually, is for non-commercial research only; the MIT license covers the library code. face_identity_extractor bills customers for this pack, so a commercial license from InsightFace is needed."
      },
      {
        "q": "How does InsightFace buffalo_l (SCRFD-10GF + ArcFace ResNet50 WebFace600K) run on Mixpeek?",
        "a": "insightface__arcface Serve app; face_identity_extractor batch calls it (FaceProcessingBatchService, strict_serve, no local fallback)."
      }
    ]
  },
  "last_verified": "2026-11-05",
  "sources": [
    "https://github.com/deepinsight/insightface/blob/master/model_zoo/README.md",
    "https://github.com/deepinsight/insightface/blob/master/python-package/README.md",
    "https://arxiv.org/abs/1801.07698",
    "https://arxiv.org/abs/2105.04714",
    "server/engine/inference/insightface/routes.py",
    "server/engine/inference/insightface/scrfd/services.py",
    "server/engine/inference/insightface/arcface/services.py",
    "server/engine/extractors/face_identity_extractor/v1/definition.py",
    "server/engine/extractors/face_identity_extractor/v1/pipeline.py",
    "server/engine/extractors/face_identity_extractor/v1/processors/processors.py"
  ],
  "registry": {
    "insightface__arcface": {
      "deployment": "insightface__arcface",
      "service_id": "insightface/arcface",
      "routes": "engine/inference/insightface/routes.py",
      "route_prefix": "/face",
      "enabled": true,
      "description": "InsightFace ArcFace R100 face embedding service. Executes complete SOTA face recognition pipeline: SCRFD detection → 5-point affine alignment → ArcFace 512-d embedding generation. 99.83% accuracy on LFW benchmark.",
      "capabilities": {
        "category": "embedding",
        "modality_input": [
          "image"
        ],
        "modality_output": "vector",
        "supports_batch": true,
        "reranking_mode": null
      },
      "vector_config": {
        "dimensions": 512,
        "distance": "cosine",
        "vector_type": "dense"
      },
      "serve": {
        "min_replicas": 0,
        "max_replicas": 4,
        "idle_timeout_s": 600,
        "target_ongoing_requests": 8,
        "num_cpus": 0.5,
        "num_gpus": {
          "expr": "hardware_config.gpu_fraction if hardware_config.gpu_concurrency > 0 else 0"
        },
        "memory_gb": 2,
        "pinnable": false,
        "placement": "worker",
        "enabled": true
      },
      "interactive_twin": false,
      "interactive_serve": null,
      "schemas": {
        "input": "engine.inference.insightface.models.FaceEmbedRequest",
        "output": "engine.inference.insightface.models.FaceEmbedResponse",
        "config": null
      }
    }
  },
  "templates": [],
  "usage": {
    "namespaces": null,
    "collections": null,
    "uncounted": true,
    "templates": 0,
    "counted_at": null,
    "source": null
  },
  "facets": {
    "status": "served",
    "category": "face",
    "family": "insightface",
    "license": "InsightFace pretrained models, non-commercial research only (library code MIT)",
    "commercial_use": false,
    "modalities_in": [
      "image",
      "video"
    ],
    "modalities_out": [
      "boxes",
      "keypoints",
      "vector"
    ],
    "use_cases": [
      "face-search"
    ],
    "recommended": false,
    "release_date": null,
    "params_millions": null,
    "size_on_disk_mb": 326,
    "vector_dims": 512,
    "vector_distance": "cosine",
    "weights_format": "onnx",
    "formats": [],
    "loadable_today": false,
    "trust_remote_code": false,
    "cpu_fallback": true,
    "num_gpus": null,
    "memory_gb": 2,
    "min_replicas": 0,
    "serving_shape": "both",
    "served": true,
    "ref_schemes": [
      "github"
    ],
    "has_evidence": false,
    "cost_measured": false,
    "template_count": 0,
    "usage_namespaces": null,
    "usage_collections": null,
    "usage_uncounted": true,
    "has_demo": false,
    "rank_score": 3
  },
  "rank_score": 3
}