Skip to main content

Lightweight Nudity Detection

Project description

Looking for contributors/ maintainers for this repo: I have become busy with other stuff in the last years, still trying to maintain this repo as it is the current best OSS option for nudity detection, Looking for interested mainttainer, who can add/ work on more features for this repo (with my help of course)

NudeNet: lightweight Nudity detection

https://nudenet.notai.tech/ in-browser demo (the detector is run client side, i.e: in your browser, images are not sent to a server)

pip install --upgrade "nudenet>=3.4.2"
from nudenet import NudeDetector
detector = NudeDetector()
# the 320n model included with the package will be used

detector.detect('image.jpg') # Returns list of detections

detector.detect_batch(['image_1.jpg', 'image_2.jpg']) # Returns list of [list of detections]
  • Python package example in colab

  • detect and detect_batch accept file path(s), opencv image(s), image bytes(s), open(image_path, 'rb') (buffereader) objects

Available models

Model resolution trained based on onnx link pytorch link
320n 320x320 ultralytics yolov8n link link
640m 640x640 ultralytics yolov8m link link
# To use the 640m model, download the onnx file and pass the path to the model_path argument

detector = NudeDetector(model_path="downloaded_640m.onnx path", inference_resolution=640)
  • 320n is the default model and is included in the nudenet python package by default
detection_example = [
 {'class': 'BELLY_EXPOSED',
  'score': 0.799403190612793,
  'box': [64, 182, 49, 51]},
 {'class': 'FACE_FEMALE',
  'score': 0.7881264686584473,
  'box': [82, 66, 36, 43]},
 ]
nude_detector.censor('image.jpg') # returns censored image output path

# optional censor(self, image_path, classes=[], output_path=None) classes and output_path can be passed
all_labels = [
    "FEMALE_GENITALIA_COVERED",
    "FACE_FEMALE",
    "BUTTOCKS_EXPOSED",
    "FEMALE_BREAST_EXPOSED",
    "FEMALE_GENITALIA_EXPOSED",
    "MALE_BREAST_EXPOSED",
    "ANUS_EXPOSED",
    "FEET_EXPOSED",
    "BELLY_COVERED",
    "FEET_COVERED",
    "ARMPITS_COVERED",
    "ARMPITS_EXPOSED",
    "FACE_MALE",
    "BELLY_EXPOSED",
    "MALE_GENITALIA_EXPOSED",
    "ANUS_COVERED",
    "FEMALE_BREAST_COVERED",
    "BUTTOCKS_COVERED",
]

Docker

docker run -it -p8080:8080 ghcr.io/notai-tech/nudenet:latest
curl -F f1=@"images.jpeg" "http://localhost:8080/infer"

{"prediction": [[{"class": "BELLY_EXPOSED", "score": 0.8511635065078735, "box": [71, 182, 31, 50]}, {"class": "FACE_FEMALE", "score": 0.8033977150917053, "box": [83, 69, 21, 37]}, {"class": "FEMALE_BREAST_EXPOSED", "score": 0.7963727712631226, "box": [85, 137, 24, 38]}, {"class": "FEMALE_BREAST_EXPOSED", "score": 0.7709134817123413, "box": [63, 136, 20, 37]}, {"class": "ARMPITS_EXPOSED", "score": 0.7005534172058105, "box": [60, 127, 10, 20]}, {"class": "FEMALE_GENITALIA_EXPOSED", "score": 0.6804671287536621, "box": [81, 241, 14, 24]}]], "success": true}

Some interesting projects based on NudeNet

1 - by https://github.com/w-e-w, censor extension ps://github.com/notAI-tech/NudeNet/issues/131

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

nudenet-3.4.2.tar.gz (10.6 MB view hashes)

Uploaded Source

Built Distribution

nudenet-3.4.2-py3-none-any.whl (10.6 MB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page