Although artificial intelligence is rapidly advancing and image processing technology is continuously upgrading, background removal has always been one of the most challenging tasks in image processing.
Currently, the best performer in background removal is still remove.bg, which does it quickly and effectively.
Now, there is finally an open-source software with background removal capabilities comparable to remove.bg. The RMBG 2.0 released by the Bria team delivers astonishing results.
Compared to its predecessor RMBG 1.4, the new version has made significant improvements in several core areas. RMBG 2.0 integrates advanced semantic segmentation, enhanced edge detection technology, and maintains stronger consistency when processing similar images, especially excelling in handling complex backgrounds.
This next-generation open-source background removal tool, based on the BiRefNet architecture, achieves an astonishing overall success rate of 90.14%. Particularly when dealing with realistic images, its accuracy reaches as high as 92%.
Even in scenarios with complex backgrounds, the tool maintains an excellent performance of 87%, showcasing its strong technical capabilities.
AIbase uploaded a picture of a red panda and found that RMBG 2.0 can cleanly remove the furry subject, with the edges of the fur clearly visible:
For images where the subject and background are highly integrated and the edges are hard to distinguish, it can easily separate the foreground from the background, completing high-quality background removal:
Removing backgrounds from anime-style images is also very easy, especially with the hair strands being cut out very cleanly:
The most important point is that it allows you to download high-resolution images for free. Essentially, the size of the uploaded image is the same as the size of the downloaded image. In contrast, remove.bg requires payment for high-resolution downloads, and the prices are not cheap.
However, compared to remove.bg, a drawback of RMBG 2.0 is that it does not provide brush or secondary editing features. This means that for images with many subjects, it may not be able to achieve a clean removal, and there is no way to edit or modify:
But considering that it's free, the background removal effect of RMBG 2.0 is quite satisfactory.
This model can be applied in the following scenarios:
E-commerce product visualization
Marketing content automation
Batch image processing
Game asset development
Website construction automation
If you're interested, click here to experience it: https://huggingface.co/spaces/briaai/BRIA-RMBG-2.0