{ "id": 62606, "title": "References that can help you ...", "url": "/competitions/airbus-ship-detection/discussion/62606", "author_name": "Paulo Pinto", "post_date": "2018-08-03T21:57:53.730000", "votes": 29, "comment_count": 10, "views": 0, "content": "

https://www.kaggle.com/rhammell/ships-in-satellite-imagery

\n\n

https://www.researchgate.net/publication/324012764_Automatic_Ship_Classification_from_Optical_Aerial_Images_with_Convolutional_Neural_Networks

", "messages": [ { "id": 366052, "postDate": "2018-08-03T21:57:53.730Z", "content": "

https://www.kaggle.com/rhammell/ships-in-satellite-imagery

\n\n

https://www.researchgate.net/publication/324012764_Automatic_Ship_Classification_from_Optical_Aerial_Images_with_Convolutional_Neural_Networks

", "rawMarkdown": "https://www.kaggle.com/rhammell/ships-in-satellite-imagery\n\nhttps://www.researchgate.net/publication/324012764_Automatic_Ship_Classification_from_Optical_Aerial_Images_with_Convolutional_Neural_Networks\n", "votes": 28 }, { "id": 386322, "postDate": "2018-09-12T15:52:46.300Z", "content": "

New article on ship detection from a company specialized in remote sensing AI: \nhttps://medium.com/earthcube-stories/how-hard-it-is-for-an-ai-to-detect-ships-on-satellite-images-7265e34aadf0

", "rawMarkdown": "New article on ship detection from a company specialized in remote sensing AI: \nhttps://medium.com/earthcube-stories/how-hard-it-is-for-an-ai-to-detect-ships-on-satellite-images-7265e34aadf0", "votes": 1, "replies": [ { "id": 386363, "postDate": "2018-09-12T17:52:57.813Z", "content": "

Thanks for sharing!

", "rawMarkdown": "Thanks for sharing!" } ] }, { "id": 368948, "postDate": "2018-08-11T11:27:33.523Z", "content": "

You can also look for top solutions from previous (and similar) competitions. One example is 1st place solution in DSB'18.

", "rawMarkdown": "You can also look for top solutions from previous (and similar) competitions. One example is [1st place solution](https://www.kaggle.com/c/data-science-bowl-2018/discussion/54741) in DSB'18.", "votes": 1, "replies": [ { "id": 380114, "postDate": "2018-09-01T18:30:44.893Z", "content": "

Indeed, the first place of this year's DSB competition has a lot of neat tricks. :)

", "rawMarkdown": "Indeed, the first place of this year's DSB competition has a lot of neat tricks. :)" } ] }, { "id": 420035, "postDate": "2018-11-13T00:48:01.473Z", "content": "

Probably a little too late for this competition, but a paper just showed up on arxiv featuring some techniques for domain adaptation transfer learning for image segmentation using pseudo labeling (proxy labeling) hard source mining and easy target mining.

\n\n

https://arxiv.org/pdf/1811.03542.pdf

", "rawMarkdown": "Probably a little too late for this competition, but a paper just showed up on arxiv featuring some techniques for domain adaptation transfer learning for image segmentation using pseudo labeling (proxy labeling) hard source mining and easy target mining.\n\nhttps://arxiv.org/pdf/1811.03542.pdf " }, { "id": 416061, "postDate": "2018-11-06T05:29:27.763Z", "content": "

This paper from August this year seems very relevant. Has anyone tried to implement it?

\n\n

Multiscale Rotated Bounding Box-Based Deep Learning Method for Detecting Ship Targets in Remote Sensing Images

\n\n

Update: Looks like this paper is an improvement on the DRBox algorithm which is referenced here.

", "rawMarkdown": "This paper from August this year seems very relevant. Has anyone tried to implement it?\n\n[Multiscale Rotated Bounding Box-Based Deep Learning Method for Detecting Ship Targets in Remote Sensing Images](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6112046/)\n\nUpdate: Looks like this paper is an improvement on the DRBox algorithm which is referenced [here](https://www.kaggle.com/c/airbus-ship-detection/discussion/62273#389671)." }, { "id": 372548, "postDate": "2018-08-19T17:04:03.613Z", "content": "

Stanford Detection & Segmentation lecture that covers U-net and Mask RCNN architectures https://youtu.be/nDPWywWRIRo?t=16m58s\nSlides to accompany the video http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture11.pdf

", "rawMarkdown": "Stanford Detection & Segmentation lecture that covers U-net and Mask RCNN architectures https://youtu.be/nDPWywWRIRo?t=16m58s\nSlides to accompany the video http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture11.pdf" }, { "id": 371727, "postDate": "2018-08-17T14:33:04.480Z", "content": "

Excellent page describing the differences between Conv2DTranspose Vs Upsampling + Conv2D\nhttps://distill.pub/2016/deconv-checkerboard/

", "rawMarkdown": "Excellent page describing the differences between Conv2DTranspose Vs Upsampling + Conv2D\nhttps://distill.pub/2016/deconv-checkerboard/" }, { "id": 366762, "postDate": "2018-08-06T12:57:50.853Z", "content": "

And the paper introducing U-net:\nhttps://pdfs.semanticscholar.org/0704/5f87709d0b7b998794e9fa912c0aba912281.pdf

", "rawMarkdown": "And the paper introducing U-net:\nhttps://pdfs.semanticscholar.org/0704/5f87709d0b7b998794e9fa912c0aba912281.pdf" }, { "id": 416062, "postDate": "2018-11-06T05:31:38.890Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 386322, "author_name": "Alexis Letulier", "author_url": "", "post_date": "2018-09-12T15:52:46.300000", "content": "

New article on ship detection from a company specialized in remote sensing AI: \nhttps://medium.com/earthcube-stories/how-hard-it-is-for-an-ai-to-detect-ships-on-satellite-images-7265e34aadf0

", "votes": 1, "replies": [ { "id": 386363, "author_name": "Yassine Alouini", "author_url": "", "post_date": "2018-09-12T17:52:57.813000", "content": "

Thanks for sharing!

", "votes": 0, "replies": [] } ] }, { "id": 368948, "author_name": "kamil", "author_url": "", "post_date": "2018-08-11T11:27:33.523000", "content": "

You can also look for top solutions from previous (and similar) competitions. One example is 1st place solution in DSB'18.

", "votes": 1, "replies": [ { "id": 380114, "author_name": "Yassine Alouini", "author_url": "", "post_date": "2018-09-01T18:30:44.893000", "content": "

Indeed, the first place of this year's DSB competition has a lot of neat tricks. :)

", "votes": 0, "replies": [] } ] }, { "id": 420035, "author_name": "Paul Johnson", "author_url": "", "post_date": "2018-11-13T00:48:01.473000", "content": "

Probably a little too late for this competition, but a paper just showed up on arxiv featuring some techniques for domain adaptation transfer learning for image segmentation using pseudo labeling (proxy labeling) hard source mining and easy target mining.

\n\n

https://arxiv.org/pdf/1811.03542.pdf

", "votes": 0, "replies": [] }, { "id": 416061, "author_name": "gauss256", "author_url": "", "post_date": "2018-11-06T05:29:27.763000", "content": "

This paper from August this year seems very relevant. Has anyone tried to implement it?

\n\n

Multiscale Rotated Bounding Box-Based Deep Learning Method for Detecting Ship Targets in Remote Sensing Images

\n\n

Update: Looks like this paper is an improvement on the DRBox algorithm which is referenced here.

", "votes": 0, "replies": [] }, { "id": 372548, "author_name": "Jack Vial", "author_url": "", "post_date": "2018-08-19T17:04:03.613000", "content": "

Stanford Detection & Segmentation lecture that covers U-net and Mask RCNN architectures https://youtu.be/nDPWywWRIRo?t=16m58s\nSlides to accompany the video http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture11.pdf

", "votes": 0, "replies": [] }, { "id": 371727, "author_name": "Paul Johnson", "author_url": "", "post_date": "2018-08-17T14:33:04.480000", "content": "

Excellent page describing the differences between Conv2DTranspose Vs Upsampling + Conv2D\nhttps://distill.pub/2016/deconv-checkerboard/

", "votes": 0, "replies": [] }, { "id": 366762, "author_name": "Adams", "author_url": "", "post_date": "2018-08-06T12:57:50.853000", "content": "

And the paper introducing U-net:\nhttps://pdfs.semanticscholar.org/0704/5f87709d0b7b998794e9fa912c0aba912281.pdf

", "votes": 0, "replies": [] }, { "id": 416062, "author_name": "", "author_url": "", "post_date": "2018-11-06T05:31:38.890000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "366052": "https://www.kaggle.com/rhammell/ships-in-satellite-imagery\n\nhttps://www.researchgate.net/publication/324012764_Automatic_Ship_Classification_from_Optical_Aerial_Images_with_Convolutional_Neural_Networks\n", "386322": "New article on ship detection from a company specialized in remote sensing AI: \nhttps://medium.com/earthcube-stories/how-hard-it-is-for-an-ai-to-detect-ships-on-satellite-images-7265e34aadf0", "368948": "You can also look for top solutions from previous (and similar) competitions. One example is [1st place solution](https://www.kaggle.com/c/data-science-bowl-2018/discussion/54741) in DSB'18.", "420035": "Probably a little too late for this competition, but a paper just showed up on arxiv featuring some techniques for domain adaptation transfer learning for image segmentation using pseudo labeling (proxy labeling) hard source mining and easy target mining.\n\nhttps://arxiv.org/pdf/1811.03542.pdf ", "416061": "This paper from August this year seems very relevant. Has anyone tried to implement it?\n\n[Multiscale Rotated Bounding Box-Based Deep Learning Method for Detecting Ship Targets in Remote Sensing Images](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6112046/)\n\nUpdate: Looks like this paper is an improvement on the DRBox algorithm which is referenced [here](https://www.kaggle.com/c/airbus-ship-detection/discussion/62273#389671).", "372548": "Stanford Detection & Segmentation lecture that covers U-net and Mask RCNN architectures https://youtu.be/nDPWywWRIRo?t=16m58s\nSlides to accompany the video http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture11.pdf", "371727": "Excellent page describing the differences between Conv2DTranspose Vs Upsampling + Conv2D\nhttps://distill.pub/2016/deconv-checkerboard/", "366762": "And the paper introducing U-net:\nhttps://pdfs.semanticscholar.org/0704/5f87709d0b7b998794e9fa912c0aba912281.pdf", "416062": "" } }