DEV Community

Cover image for BiRefNet State Of The Art Newest Very Best Background Batch Remover APP
Furkan Gözükara
Furkan Gözükara

Posted on

BiRefNet State Of The Art Newest Very Best Background Batch Remover APP

BiRefNet State Of The Art Newest Very Best Background Batch Remover APP. BiRefNet is the very best state of the art newest published background removal open source model. It is way better than BRIA RMBG V1.4 that we knew as best.

Official repo : https://github.com/ZhengPeng7/BiRefNet

Download APP and installers from : https://www.patreon.com/posts/109913645

Image description

Hugging Face Demo : https://huggingface.co/spaces/ZhengPeng7/BiRefNet_demo

I have developed a very advanced Gradio APP for this with full proper file saving and batch processing. Also my version removes BG and saves as transparent background.

The APP uses huge VRAM for high resolution images. However it is still working uber fast even though using shared VRAM. So make sure that you have high RAM or set virtual RAM.

Click below to see how to set virtual RAM on Windows.

https://www.windowscentral.com/how-change-virtual-memory-size-windows-10

On Massed Compute A6000 GPU (31 cents per hour) you can very fast remove even very high res images backgrounds.

Currently we have 1 click installers for RunPod, Massed Compute, Kaggle and Windows.

Windows Requirements

Python 3.10, FFmpeg, Cuda 11.8, C++ tools and Git

If it doesn’t work make sure to below tutorial and install everything exactly as shown in this below tutorial

https://youtu.be/-NjNy7afOQ0

How To Use On Windows

Just extract files into like c:/BiRefNet_v1

Double click Windows_Install.bat file and it will generate a isolated virtual environment and install requirements

It will automatically download models into your Hugging Face cache (best model under 1 GB)

Then start and use the Gradio APP with Windows_Start_App.bat

Cloud How To Use

Massed Compute, RunPod has instructions txt files. Follow them

Kaggle has all the instructions 1 by 1

On Kaggle set resolution 1024x1024 or you will get out of memory error

Image description

Top comments (0)