Summary of Safari:adaptive Sequence Transformer For Weakly Supervised Referring Expression Segmentation, by Sayan Nag et al.
SafaRi:Adaptive Sequence Transformer for Weakly Supervised Referring Expression Segmentation
by Sayan Nag, Koustava Goswami, Srikrishna Karanam
First submitted to arxiv on: 2 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG); Multimedia (cs.MM)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes a weakly-supervised bootstrapping architecture for Referring Expression Segmentation (RES), which aims to provide a segmentation mask of the target object in an image referred to by the text. The existing methods require large-scale mask annotations, but this approach only needs a fraction of both mask and box annotations for training. The proposed Cross-modal Fusion with Attention Consistency module improves image-text region-level alignment and spatial localization of the target object. A novel Mask Validity Filtering routine is introduced for automatic pseudo-labeling of unlabeled samples. Experimental results show that the model SafaRi achieves competitive performance to fully-supervised methods, even in a zero-shot setting. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us better understand how computers can identify objects in images when we describe them using words. Right now, it’s hard for computers to do this without having lots of information about what’s in the image. The researchers came up with a new way to teach computers to do this task, even if they don’t have much information. They call this method “weakly-supervised bootstrapping” and it uses special tricks to make sure the computer is learning the right things from the images. |
Keywords
» Artificial intelligence » Alignment » Attention » Bootstrapping » Mask » Supervised » Zero shot
