I'm looking to create a browser extension that blurs images and videos if a full human body is detected. The idea is straightforward: when a human is detected, apply a CSS blur effect; if not, then continue as normal. I need a lightweight, fast model that runs client-side without much latency. Larger models like YOLO are too overwhelming for this purpose. I'm also curious about which runtimes are the most efficient for this task, such as TensorFlow.js or ONNX Runtime Web. Additionally, I'd like some guidance on running the model without triggering CORS issues or other errors that could hinder its functionality. I've tried using AI tools for recommendations, but I want to hear from anyone with practical insights or alternatives that AI might not cover.
2 Answers
Honestly, your question sounds a bit confusing to me! But if you're focused on making a browser extension that detects humans in images for blurring, that's pretty clear.
You might want to consider blurring on the backend instead of the frontend. If you're using JavaScript to blur images on the user side, the original image can still be accessed, which might not be ideal for privacy.
Since the goal is to blur images on the fly across various pages, there’s no backend needed. The blur will only apply if a human is detected, so the original won't be a problem in this case.
I get where you're coming from; the question may seem a bit scattered but the main goal is to develop an extension that identifies human figures in images and applies a blur for each one detected. The reasoning behind it is still a bit unclear to me, though.