{ "id": 412316, "title": "SOTA and Popular Segment method: OneFormer and UNet", "url": "/competitions/hubmap-hacking-the-human-vasculature/discussion/412316", "author_name": "Dewei Chen", "post_date": "2023-05-23T07:29:36.115000", "votes": 30, "comment_count": 4, "views": 0, "content": "

1. OneFormer: One Transformer to Rule Universal Image Segmentation

\n

OneFormer is a method based on transformer, and achieve the SOTA in 2023 CVPR
\nthe original paper: https://arxiv.org/pdf/2211.06220v2.pdf
\nan implementation in github: https://github.com/SHI-Labs/OneFormer

\n

.

\n

2. U-Net: Convolutional Networks for Biomedical Image Segmentation

\n

The U-Net is the most popular segmentation method in bio relation topic.
\nthe original paper: https://arxiv.org/pdf/1505.04597v1.pdf
\nan implementation in github: https://github.com/milesial/Pytorch-UNet or https://github.com/open-mmlab/mmsegmentation

", "messages": [ { "id": 2270502, "postDate": "2023-05-23T07:29:36.117Z", "content": "

1. OneFormer: One Transformer to Rule Universal Image Segmentation

\n

OneFormer is a method based on transformer, and achieve the SOTA in 2023 CVPR
\nthe original paper: https://arxiv.org/pdf/2211.06220v2.pdf
\nan implementation in github: https://github.com/SHI-Labs/OneFormer

\n

.

\n

2. U-Net: Convolutional Networks for Biomedical Image Segmentation

\n

The U-Net is the most popular segmentation method in bio relation topic.
\nthe original paper: https://arxiv.org/pdf/1505.04597v1.pdf
\nan implementation in github: https://github.com/milesial/Pytorch-UNet or https://github.com/open-mmlab/mmsegmentation

", "rawMarkdown": "##1. **OneFormer: One Transformer to Rule Universal Image Segmentation**\nOneFormer is a method based on transformer, and achieve the SOTA in 2023 CVPR\nthe original paper: https://arxiv.org/pdf/2211.06220v2.pdf\nan implementation in github: https://github.com/SHI-Labs/OneFormer\n\n\n.\n\n##2. **U-Net: Convolutional Networks for Biomedical Image Segmentation**\nThe U-Net is the most popular segmentation method in bio relation topic.\nthe original paper: https://arxiv.org/pdf/1505.04597v1.pdf\nan implementation in github: https://github.com/milesial/Pytorch-UNet or https://github.com/open-mmlab/mmsegmentation", "votes": 30 }, { "id": 2312421, "postDate": "2023-06-21T21:54:39.380Z", "content": "

Do you think OneFormer is the best model for segmentation out there at the moment? U-Net is quite old now

", "rawMarkdown": "Do you think OneFormer is the best model for segmentation out there at the moment? U-Net is quite old now", "votes": 1, "replies": [ { "id": 2320306, "postDate": "2023-06-27T17:15:39.950Z", "content": "

It seems that OneFormer is the state of the art (as of June 2023) at least for panoptic segmentation as stated by this link.

\n

\"\"

", "rawMarkdown": "It seems that `OneFormer` is the state of the art (as of June 2023) at least for panoptic segmentation as stated by this [link](https://paperswithcode.com/sota/panoptic-segmentation-on-ade20k-val). \n\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F172860%2F887879f836d3b2a2d33c2e488fa536b6%2FScreenshot%20from%202023-06-27%2019-14-56.png?generation=1687886132706239&alt=media)\n" } ] }, { "id": 2320296, "postDate": "2023-06-27T17:07:29.127Z", "content": "

Thanks for the OneFormer link, first time I hear about it. What about SegFormer?
\nFYI: I found a nice summary about OneFormer here.

", "rawMarkdown": "Thanks for the `OneFormer` link, first time I hear about it. What about `SegFormer`?\nFYI: I found a nice summary about `OneFormer` [here](https://huggingface.co/blog/mask2former)." }, { "id": 2274551, "postDate": "2023-05-26T05:13:21.513Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 2312421, "author_name": "Samuel Cortinhas", "author_url": "", "post_date": "2023-06-21T21:54:39.380000", "content": "

Do you think OneFormer is the best model for segmentation out there at the moment? U-Net is quite old now

", "votes": 1, "replies": [ { "id": 2320306, "author_name": "Yassine Alouini", "author_url": "", "post_date": "2023-06-27T17:15:39.950000", "content": "

It seems that OneFormer is the state of the art (as of June 2023) at least for panoptic segmentation as stated by this link.

\n

\"\"

", "votes": 0, "replies": [] } ] }, { "id": 2320296, "author_name": "Yassine Alouini", "author_url": "", "post_date": "2023-06-27T17:07:29.127000", "content": "

Thanks for the OneFormer link, first time I hear about it. What about SegFormer?
\nFYI: I found a nice summary about OneFormer here.

", "votes": 0, "replies": [] }, { "id": 2274551, "author_name": "", "author_url": "", "post_date": "2023-05-26T05:13:21.513000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2270502": "##1. **OneFormer: One Transformer to Rule Universal Image Segmentation**\nOneFormer is a method based on transformer, and achieve the SOTA in 2023 CVPR\nthe original paper: https://arxiv.org/pdf/2211.06220v2.pdf\nan implementation in github: https://github.com/SHI-Labs/OneFormer\n\n\n.\n\n##2. **U-Net: Convolutional Networks for Biomedical Image Segmentation**\nThe U-Net is the most popular segmentation method in bio relation topic.\nthe original paper: https://arxiv.org/pdf/1505.04597v1.pdf\nan implementation in github: https://github.com/milesial/Pytorch-UNet or https://github.com/open-mmlab/mmsegmentation", "2312421": "Do you think OneFormer is the best model for segmentation out there at the moment? U-Net is quite old now", "2320296": "Thanks for the `OneFormer` link, first time I hear about it. What about `SegFormer`?\nFYI: I found a nice summary about `OneFormer` [here](https://huggingface.co/blog/mask2former).", "2274551": "" } }