MediaPipe BlazeFace (Short-Range)

Lightweight single-shot face detector (blaze_face_short_range.tflite, 230 KB, FP16 weights) optimized for selfie and front-facing smartphone camera or webcam images at short range (within ≈2 meters), packaged for the MediaPipe Face Detector (FaceDetector) task and LiteRT.

Documentation: MediaPipe Face Detector Docs

Samples: Android · iOS · Python Colab · Web · Flutter

Try live: Web Demo

BlazeFace (short-range) uses a Single Shot MultiBox Detector (SSD) architecture with a lightweight MobileNetV1/V2-inspired feature extractor tailored for mobile GPU and CPU (XNNPACK) execution.

For each detected face, the MediaPipe FaceDetector task decodes the raw anchor predictions and applies Non-Maximum Suppression (NMS) to return:

  • Axis-aligned bounding box (origin_x, origin_y, width, height in pixels)
  • 6 approximate 2D facial keypoints normalized to [0.0, 1.0] (left_eye, right_eye, nose_tip, mouth, left_eye_tragion, right_eye_tragion from the observer's perspective)
  • Detection confidence score in [0.0, 1.0]

Files

File Precision Size SHA-256
blaze_face_short_range.tflite float16 229.7 KB (229,746 B) b4578f35940bf5a1a655214a1cce5cab13eba73c1297cd78e1a04c2380b0152f

Companion Files in Repo Root

File Role Size SHA-256
MediaPipe BlazeFace Model Card (Short Range).pdf Upstream Google model card (PDF) 434.4 KB (434,425 B) cd335c06fc0de7807cd815a0777a697932598bcdb28fa98adaaabf847485f758

Download

pip install -U huggingface_hub
hf download litert-community/MediaPipe-FaceDetector-BlazeFace-Short-Range blaze_face_short_range.tflite --local-dir .

Inputs, Outputs, and Model Specifications

blaze_face_short_range.tflite

  • Embedded TFLITE_METADATA: name Short Range Face Detection · version 1 · author MediaPipe
Direction Tensor Shape Type Notes
Input input [1, 128, 128, 3] float32 image — Input image to be detected — Normalization: mean=[127.5], std=[127.5]
Output regressors [1, 896, 16] float32 raw boxes/keypoints — Undecoded face bboxes location and keypoints
Output classificators [1, 896, 1] float32 scores — Scores of the detected bboxes.

Use with MediaPipe Tasks

Python Quickstart

pip install -U mediapipe huggingface_hub
import mediapipe as mp
from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="litert-community/MediaPipe-FaceDetector-BlazeFace-Short-Range",
    filename="blaze_face_short_range.tflite",
)

BaseOptions = mp.tasks.BaseOptions
FaceDetector = mp.tasks.vision.FaceDetector
FaceDetectorOptions = mp.tasks.vision.FaceDetectorOptions
VisionRunningMode = mp.tasks.vision.RunningMode

options = FaceDetectorOptions(
    base_options=BaseOptions(model_asset_path=model_path),
    running_mode=VisionRunningMode.IMAGE,
    min_detection_confidence=0.5,
)

with FaceDetector.create_from_options(options) as runner:
    mp_image = mp.Image.create_from_file("input.jpg")
    result = runner.detect(mp_image)
    for detection in result.detections:
        bbox = detection.bounding_box
        score = detection.categories[0].score
        print(f"Face score={score:.3f} box=({bbox.origin_x}, {bbox.origin_y}, {bbox.width}, {bbox.height})")

Cross-Platform Guides & Samples

Latency Benchmarks

End-to-end MediaPipe Task pipeline average latency reported by the Face Detector documentation:

Model Variant Device CPU Latency GPU Latency
BlazeFace (short-range) Pixel 6 2.94 ms 7.41 ms

Intended Use, Limitations, and Fairness

Summarized from the bundled MediaPipe BlazeFace Model Card (Short Range).pdf (Google, June 9, 2021):

  • Intended applications: Detecting prominently displayed faces in smartphone camera or webcam images and video frames for live perception pipelines, mobile augmented reality (AR), entertainment, and assistive experiences.
  • Operating envelope: Designed for faces within ≈2 meters of the camera, where face bounding box sides are at least 20% of the corresponding image dimensions, at least 70% of the face lies within the frame, roll/pitch angles are within ±45°, and yaw is within ±90°.
  • Out-of-scope uses:
    • Any form of surveillance or identity recognition is explicitly out of scope and not enabled by this technology.
    • Human life-critical decisions.
    • Counting people in crowds, detecting faces > 2 meters from the camera, or detecting faces turned away from the camera.
  • Known limitations: Outputs up to a fixed cap of detections even when more people are in view; extreme lighting, motion blur, sensor noise, or face occlusion can degrade accuracy and increase inter-frame jitter.
  • Fairness evaluation (summary): Evaluated on consented in-the-wild smartphone AR images across 17 UN geographic subregions (global average recall 99.1%, range 92.5%–100%; global average precision 95.1%, range 94.9%–95.2%), annotator-perceived gender presentation (average recall 98.4%, precision 99.9%), and 5 Fitzpatrick-aligned skin-tone groups (average recall 98.1%, range 94.7%–100%; average precision 99.7%, range 98.6%–100%). Refer to the bundled PDF model card for full methodology and definitions.

Provenance

Every model binary in this repository is byte-identical (unmodified FlatBuffer weights and embedded TFLITE_METADATA bytes) to the official artifact served by Google from storage.googleapis.com/mediapipe-models.

Repo File Upstream URL GCS Generation Upstream Last-Modified Pinned Bucket Version
blaze_face_short_range.tflite blaze_face_short_range.tflite 1682480001393569 Wed, 26 Apr 2023 03:33:21 GMT /1/

License and Citation

  • License: apache-2.0 — Stated as Apache License, Version 2.0 in the bundled upstream Google model-card PDF.

Citation

@inproceedings{bazarevsky2019blazeface,
  title     = {BlazeFace: Sub-millisecond Neural Face Detection on Mobile GPUs},
  author    = {Bazarevsky, Valentin and Kartynnik, Yury and Vakunov, Andrey and Raveendran, Karthik and Grundmann, Matthias},
  booktitle = {CVPR Workshop on Computer Vision for Augmented and Virtual Reality},
  year      = {2019},
  url       = {https://arxiv.org/abs/1907.05047}
}
Downloads last month
-
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Collection including litert-community/MediaPipe-FaceDetector-BlazeFace-Short-Range

Paper for litert-community/MediaPipe-FaceDetector-BlazeFace-Short-Range