real time

[luantunez]

Hello and thank you for sharing your work!
I do not understand entirely how does the real time model for skin segmentation work.
Do you have some instructions regarding, the input, the output, and the purpose?
Thank you, Lucia

[WillBrennan]

Hi there;
The steps for recreating the skin segmentation code is in the readme. First start by creating your enviroment, then follow the section on running the pretrained projects.

This code was created as a machine-learning version of the original SkinDetector project. That project was created to show how multi-colour space thresholding is used in traditional image processing to create segmentation masks.

[luantunez]

Thank you very much for your quick response. I have run your code for an image with the FCNResNet101 model. It works great!
But what I do not fully understand is the BiSeNetV2, for the real-time models. What does it mean being real time? Does it also process an image, or a video instead? What is the difference between both models?
Thank you, Lucia

[WillBrennan]

On a HD image; BiSeNet can run at 60fps while ResNet101 can barely handle 1fps on a 1070. This code is only written for single images, but you can run BiSeNet in "real-time"

[luantunez]

Perfect! Thank you!