![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12870466%2Fbce24abfdaff76a4d2d46bf1c6ddfe1b%2FScreenshot%20from%202024-01-13%2001-16-53.png?generation=1705105117647384&alt=media\")
\n
![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F12870466%2F7c85b24461c5786dcd5ed6d2f2da7deb%2FScreenshot%20from%202024-01-13%2001-18-21.png?generation=1705105139867468&alt=media\")
I stumbled upon this video, which is a presentation of the paper MaxViT-UNet: Multi-Axis Attention for Medical Image Segmentation,
\nhttps://www.youtube.com/watch?v=f04TQyIcd00
\nThe author describes his motivations and architecture details in this talk, which can help kaggler who are new to the field and who find hard going through a paper end-to-end without any guidance:
\nThis will serve as a good intro to Multi-Axis Attention as well, which is sota in Vision transformers
\n
I stumbled upon this video, which is a presentation of the paper MaxViT-UNet: Multi-Axis Attention for Medical Image Segmentation,
\nhttps://www.youtube.com/watch?v=f04TQyIcd00
\nThe author describes his motivations and architecture details in this talk, which can help kaggler who are new to the field and who find hard going through a paper end-to-end without any guidance:
\nThis will serve as a good intro to Multi-Axis Attention as well, which is sota in Vision transformers
\n
Thanks for your sharing!😀
", "rawMarkdown": "Thanks for your sharing!😀", "votes": null }, { "id": "2602245", "postDate": "01/15/2024 02:44:57", "content": "In my experiments, the vit based method did show great potential, but the OOM problem is really difficult to solve.ðŸ˜
", "rawMarkdown": "In my experiments, the vit based method did show great potential, but the OOM problem is really difficult to solve.ðŸ˜", "votes": null }, { "id": "2602759", "postDate": "01/15/2024 10:17:19", "content": "hey @tanxxx , Are you talking about the inference oom or simply training ? if it is during inference, is it the cpu or gpu that runs out ? I don’t really have problems like that atm so I would like to know more if that’s ok to share
", "rawMarkdown": "hey @tanxxx , Are you talking about the inference oom or simply training ? if it is during inference, is it the cpu or gpu that runs out ? I don’t really have problems like that atm so I would like to know more if that’s ok to share", "votes": null }, { "id": "2602780", "postDate": "01/15/2024 10:37:59", "content": "I haven't check out the problem when simple inference, i can infer with training data in kaggle notebook, but while submit to kaggle, \"Notebook Threw Exception\" happenðŸ˜
", "rawMarkdown": "I haven't check out the problem when simple inference, i can infer with training data in kaggle notebook, but while submit to kaggle, \"Notebook Threw Exception\" happenðŸ˜", "votes": null }, { "id": "2602813", "postDate": "01/15/2024 11:01:06", "content": "Can you infer with the entier training data ? if so, there shouldn't be a problem with the test data as it only has 1500 tiff files, meaning simply being able to infer on kidney 2 should tell you if you have enough memory
", "rawMarkdown": "Can you infer with the entier training data ? if so, there shouldn't be a problem with the test data as it only has 1500 tiff files, meaning simply being able to infer on kidney 2 should tell you if you have enough memory", "votes": null }, { "id": "2603687", "postDate": "01/16/2024 01:11:07", "content": "Yes, i can infer with the entire kidney 2, but existing some error while submission😷
", "rawMarkdown": "Yes, i can infer with the entire kidney 2, but existing some error while submission😷", "votes": null } ], "comments": [ { "id": 2599781, "author_name": "gentlezdh", "author_url": "", "post_date": "01/13/2024 06:20:22", "content": "Thanks for your sharing!😀
", "votes": null, "replies": [] }, { "id": 2602245, "author_name": "tanxxx", "author_url": "", "post_date": "01/15/2024 02:44:57", "content": "In my experiments, the vit based method did show great potential, but the OOM problem is really difficult to solve.ðŸ˜
", "votes": null, "replies": [ { "id": 2602759, "author_name": "janmpia", "author_url": "", "post_date": "01/15/2024 10:17:19", "content": "hey @tanxxx , Are you talking about the inference oom or simply training ? if it is during inference, is it the cpu or gpu that runs out ? I don’t really have problems like that atm so I would like to know more if that’s ok to share
", "votes": null, "replies": [ { "id": 2602780, "author_name": "tanxxx", "author_url": "", "post_date": "01/15/2024 10:37:59", "content": "I haven't check out the problem when simple inference, i can infer with training data in kaggle notebook, but while submit to kaggle, \"Notebook Threw Exception\" happenðŸ˜
", "votes": null, "replies": [ { "id": 2602813, "author_name": "janmpia", "author_url": "", "post_date": "01/15/2024 11:01:06", "content": "Can you infer with the entier training data ? if so, there shouldn't be a problem with the test data as it only has 1500 tiff files, meaning simply being able to infer on kidney 2 should tell you if you have enough memory
", "votes": null, "replies": [ { "id": 2603687, "author_name": "tanxxx", "author_url": "", "post_date": "01/16/2024 01:11:07", "content": "Yes, i can infer with the entire kidney 2, but existing some error while submission😷
", "votes": null, "replies": [] } ] } ] } ] } ] } ], "raw_markdown_by_id": { "2599567": "I stumbled upon this video, which is a presentation of the paper [MaxViT-UNet: Multi-Axis Attention for Medical Image Segmentation](https://arxiv.org/abs/2305.08396),\n\nhttps://www.youtube.com/watch?v=f04TQyIcd00\n\nThe author describes his motivations and architecture details in this talk, which can help kaggler who are new to the field and who find hard going through a paper end-to-end without any guidance:\nThis will serve as a good intro to Multi-Axis Attention as well, which is sota in Vision transformers\n\n\n", "2599781": "Thanks for your sharing!😀", "2602245": "In my experiments, the vit based method did show great potential, but the OOM problem is really difficult to solve.ðŸ˜", "2602759": "hey @tanxxx , Are you talking about the inference oom or simply training ? if it is during inference, is it the cpu or gpu that runs out ? I don’t really have problems like that atm so I would like to know more if that’s ok to share", "2602780": "I haven't check out the problem when simple inference, i can infer with training data in kaggle notebook, but while submit to kaggle, \"Notebook Threw Exception\" happenðŸ˜", "2602813": "Can you infer with the entier training data ? if so, there shouldn't be a problem with the test data as it only has 1500 tiff files, meaning simply being able to infer on kidney 2 should tell you if you have enough memory", "2603687": "Yes, i can infer with the entire kidney 2, but existing some error while submission😷" }, "source": "meta" }