● Covered by 1 source · 1 reportLow impact1 neutral
FeyNoBg Model and NoBg Training Library Released for Automatic Background Removal
Key points
- FeyNoBg is a new SOTA model for automatic background removal.
- It performs well on eight benchmarks, leading on four.
- NoBg is an open-source library for training background removal models.
- Both FeyNoBg and NoBg are available on Hugging Face and GitHub.
Introduction of FeyNoBg and NoBg
FeyNoBg, a new state-of-the-art model for automatic background removal, has been introduced. The model has demonstrated strong performance, achieving the best published S-measure on four out of eight benchmarks and coming within 2% of the leader on the remaining four. This release aims to provide an advanced solution for image background segmentation.
Accompanying FeyNoBg is the open-source NoBg library, which was utilized to train the model. This library is available for public use, enabling developers and researchers to either run the FeyNoBg model or train their own custom background removal models. Both the model and the library are accessible on Hugging Face and GitHub, respectively.
Technical Approach to Background Removal
Background removal algorithms predict an opacity value for each pixel in an image, making background pixels transparent, foreground pixels opaque, and boundary pixels translucent. This process requires two main capabilities: separating foreground from background and precisely tracing the foreground's boundary, known as image matting.
Traditional methods often train these skills using focused datasets, which can lead to unbalanced models. FeyNoBg addresses this by integrating both foreground recognition and boundary precision, recognizing that real-world images demand proficiency in both areas for accurate removal.
FeyNoBg's Architecture and Training
FeyNoBg is built upon the BiRefNet architecture, which naturally aligns with the goal of combining foreground localization and boundary reconstruction. BiRefNet's design assigns complementary responsibilities to different parts of the model: a localization module identifies the foreground, and a reconstruction module traces the subject's boundary.
The model processes input images through a four-stage feature extractor. Early stages capture fine local details, while later stages aggregate these details into broader image representations, or feature maps. This hierarchical processing allows the model to develop a comprehensive understanding of the image content.
Availability and Impact
The release of FeyNoBg and the NoBg library provides new open-source tools for the computer vision community. Users can download FeyNoBg from Hugging Face for immediate application or utilize the NoBg library from GitHub to develop and train their own specialized background removal solutions. This contributes to the ongoing advancement of image processing techniques.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
Primary sources
GitHub feyninc/nobgReporting from
A new state-of-the-art model, FeyNoBg, has been released for automatic background removal, demonstrating strong performance across multiple benchmarks. Alongside it, the open-source NoBg library used for its training is also available, allowing others to run FeyNoBg or develop their own models. This release provides new tools for image processing tasks requiring precise background segmentation.