Build a usable browser with NSFW & Human detection feature

[HMZahir]

Features

• [ ] Blur the whole images
• [ ] Full settings integration — detection aggressiveness, blur strength. Show stats on detection.
• [ ] Implement KahfGuard API call to block users accessing KahfGuard-blocked pages
• Productionize browser for App Store launch
• Beta trial with users to get feedback on performance & reliability

Acceptance Criteria:

• The browser should have a built-in NSFW detection feature that can identify potentially inappropriate content (nudity, violence, etc.) on web pages.
• The browser should also have a human detection feature that can identify the presence of people in images and videos.
• Upon detecting NSFW and human content, the entire webpage or image/video should be blurred by default.
• Users should have the option to disable the NSFW and human detection features if desired.
• The browser should prioritize user privacy and avoid collecting any browsing data or personal information.

[tawfiq23]

In this version, 3 on-device models are used to classify images. 1.⁠ ⁠Nsfw 2.⁠ ⁠⁠Google ML kit’s face detection model 3.⁠ ⁠⁠Gender detection

I tested the predictions on some newspaper sites. Didn’t check any nsfw sites. I found some false positive and false negative predictions for nsfw and gender detection. App’s performance needs to be improved a lot.

The Image processing flow is as follows