ifnude - nudity detection that just works

A neural nets powered library that can detect nudity in images of both real humans and drawings. It takes an image as input and tells you exactly what NSFW parts of the body are visible. Optionally, you can censor the said parts.

Installation

pip install ifnude

Example

Note: Importing ifnude for the first time will download a 139MB module to "/your/home/dir/.ifnude/", just once.

from ifnude import detect

# use mode="fast" for x3 speed with slightly lower accuracy
print(detect('/path/to/nsfw.png'))

Instead of a path, you can use a variable that contains an image loaded through cv2 (opencv) or PIL (pillow).

Output

[
  {'box': [164, 188, 246, 271], 'score': 0.8253238201141357, 'label': 'EXPOSED_BREAST_F'},
  {'box': [252, 190, 335, 270], 'score': 0.8235630989074707, 'label': 'EXPOSED_BREAST_F'}
]

Censor Video

from ifnude import detector

detector.censor_video(inp_video_path='path/to/video.mp4', out_video_path='path/to/output_video.mp4') #make sure to use forward slash (/)

Output

raw video

https://streamable.com/q8jp8c

censored video

https://streamable.com/su3yko

Credits

This is fork of NudeNet library which doesn't work anymore.