![\"enter](\"https://farm8.staticflickr.com/7309/9037121736_735a512d76_o.png\" "\\"\\"")
This looks relevant:\nObject Detection on SpaceNet
", "messages": [ { "id": "159542", "postDate": "02/02/2017 18:20:19", "content": "This looks relevant:\nObject Detection on SpaceNet
", "rawMarkdown": "This looks relevant:\n[Object Detection on SpaceNet][1]\n\n\n [1]: https://medium.com/the-downlinq/object-detection-on-spacenet-5e691961d257", "votes": null }, { "id": "159919", "postDate": "02/04/2017 20:34:25", "content": "Classification and Segmentation of Satellite Orthoimagery Using Convolutional Neural Networks - looks like a nice review, discussing many design choices and how they affect performance. Check out citations too.
\n\nScene Classification via a Gradient Boosting Random Convolutional Network Framework - this is a slightly different problem - the task is just to classify which kind of scene is in a small patch. They also do gradient boosting of several networks, which might be applicable to other network architectures, maybe?
", "rawMarkdown": "[Classification and Segmentation of Satellite Orthoimagery Using Convolutional Neural Networks][1] - looks like a nice review, discussing many design choices and how they affect performance. Check out citations too.\n\n[Scene Classification via a Gradient Boosting Random Convolutional Network Framework][2] - this is a slightly different problem - the task is just to classify which kind of scene is in a small patch. They also do gradient boosting of several networks, which might be applicable to other network architectures, maybe?\n\n\n [1]: http://www.mdpi.com/2072-4292/8/4/329/html\n [2]: https://www.researchgate.net/publication/283523609_Scene_Classification_via_a_Gradient_Boosting_Random_Convolutional_Network_Framework", "votes": null }, { "id": "160170", "postDate": "02/06/2017 12:50:28", "content": "https://github.com/nshaud/DeepNetsForEO
", "rawMarkdown": "https://github.com/nshaud/DeepNetsForEO", "votes": null }, { "id": "160270", "postDate": "02/06/2017 21:24:35", "content": "Facebook sharp mask paper referenced here https://code.facebook.com/posts/561187904071636/segmenting-and-refining-images-with-sharpmask/ also looks relevant. Interesting that their architecture is very similar to UNet.
", "rawMarkdown": "Facebook sharp mask paper referenced here https://code.facebook.com/posts/561187904071636/segmenting-and-refining-images-with-sharpmask/ also looks relevant. Interesting that their architecture is very similar to UNet.", "votes": null }, { "id": "161035", "postDate": "02/10/2017 22:09:45", "content": "I really like this paper, some ideas like Atrous Lauer and CRF for post processing may potentially be applied to our problem.
\n\n", "rawMarkdown": "I really like this paper, some ideas like Atrous Lauer and CRF for post processing may potentially be applied to our problem.\n\n[DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs][1]\n\n\n [1]: https://arxiv.org/abs/1606.00915", "votes": null }, { "id": "161087", "postDate": "02/11/2017 12:04:35", "content": "A novel spectral index to automatically extract road networks from WorldView-2 satellite imagery
\n\n", "rawMarkdown": "[A novel spectral index to automatically extract road networks from WorldView-2 satellite imagery][1]\n\n[Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery][2]\n\n [1]: http://www.sciencedirect.com/science/article/pii/S111098231400043X \"A novel spectral index to automatically extract road networks from WorldView-2 satellite imagery\"\n [2]: http://www.aari.ru/docs/pub/060804/xuh06.pdf \"Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery\"", "votes": null }, { "id": "161605", "postDate": "02/14/2017 20:14:28", "content": "Semantic Segmentation of Earth Observation Data Using Multimodal and Multi-scale Deep Networks
", "rawMarkdown": "[Semantic Segmentation of Earth Observation Data Using Multimodal and Multi-scale Deep Networks][1]\n\n\n [1]: https://arxiv.org/abs/1609.06846", "votes": null }, { "id": "161988", "postDate": "02/16/2017 18:06:11", "content": "Semantic Segmentation Of Aerial Images Wtih An Ensemble Of CNNs 2-head modified FCN + CRF
", "rawMarkdown": "[Semantic Segmentation Of Aerial Images Wtih An Ensemble Of CNNs][1] 2-head modified FCN + CRF\n\n\n [1]: http://intanto.net/publications/Marmanis_isprs16.pdf", "votes": null }, { "id": "162489", "postDate": "02/19/2017 13:00:52", "content": "http://juliandewit.github.io/kaggle-ndsb/ - some observations about using U-net for segmentation of biomedical images
", "rawMarkdown": "http://juliandewit.github.io/kaggle-ndsb/ - some observations about using U-net for segmentation of biomedical images", "votes": null }, { "id": "163430", "postDate": "02/24/2017 00:47:32", "content": "Better understanding of different bands.\n
https://www.mapbox.com/blog/putting-landsat-8-bands-to-work/
", "rawMarkdown": "Better understanding of different bands.\n![enter image description here][1]\n\nhttps://www.mapbox.com/blog/putting-landsat-8-bands-to-work/\n\n\n [1]: https://farm8.staticflickr.com/7309/9037121736_735a512d76_o.png", "votes": null }, { "id": "163565", "postDate": "02/24/2017 15:23:05", "content": "Just to note: bands in the article (and picture above) are for Landsat-8 satellite, not for Worldview-3 we have here. And our bands are quite different.
", "rawMarkdown": "Just to note: bands in the article (and picture above) are for Landsat-8 satellite, not for Worldview-3 we have here. And our bands are quite different.", "votes": null }, { "id": "163589", "postDate": "02/24/2017 17:18:17", "content": "For sure, a numeration is quite different, but one may map Landsat-8 bands onto Worldview-3 using the wavelength.
", "rawMarkdown": "For sure, a numeration is quite different, but one may map Landsat-8 bands onto Worldview-3 using the wavelength.", "votes": null }, { "id": "163821", "postDate": "02/26/2017 05:41:03", "content": "\n\nResults don't look very promising.
", "rawMarkdown": "[Dropband: A Convolutional Neural Network With Data Augmentation For Scene Classification Of VHR Satellite Images][1] \n\nResults don't look very promising.\n [1]: http://proceedings.utwente.nl/403/1/Yang-DropBand-91.pdf", "votes": null }, { "id": "164640", "postDate": "03/01/2017 21:31:03", "content": "Here is a pdf file for Worldview-3 sensor wavebands.
\n\nhttps://www.spaceimagingme.com/downloads/sensors/datasheets/DG_WorldView3_DS_2014.pdf
", "rawMarkdown": "Here is a pdf file for Worldview-3 sensor wavebands.\n\nhttps://www.spaceimagingme.com/downloads/sensors/datasheets/DG_WorldView3_DS_2014.pdf", "votes": null } ], "comments": [ { "id": 159919, "author_name": "lopuhin", "author_url": "", "post_date": "02/04/2017 20:34:25", "content": "Classification and Segmentation of Satellite Orthoimagery Using Convolutional Neural Networks - looks like a nice review, discussing many design choices and how they affect performance. Check out citations too.
\n\nScene Classification via a Gradient Boosting Random Convolutional Network Framework - this is a slightly different problem - the task is just to classify which kind of scene is in a small patch. They also do gradient boosting of several networks, which might be applicable to other network architectures, maybe?
", "votes": null, "replies": [] }, { "id": 160170, "author_name": "cubbeewan", "author_url": "", "post_date": "02/06/2017 12:50:28", "content": "https://github.com/nshaud/DeepNetsForEO
", "votes": null, "replies": [] }, { "id": 160270, "author_name": "lopuhin", "author_url": "", "post_date": "02/06/2017 21:24:35", "content": "Facebook sharp mask paper referenced here https://code.facebook.com/posts/561187904071636/segmenting-and-refining-images-with-sharpmask/ also looks relevant. Interesting that their architecture is very similar to UNet.
", "votes": null, "replies": [] }, { "id": 161035, "author_name": "iglovikov", "author_url": "", "post_date": "02/10/2017 22:09:45", "content": "I really like this paper, some ideas like Atrous Lauer and CRF for post processing may potentially be applied to our problem.
\n\n", "votes": null, "replies": [] }, { "id": 161087, "author_name": "zfturbo", "author_url": "", "post_date": "02/11/2017 12:04:35", "content": "A novel spectral index to automatically extract road networks from WorldView-2 satellite imagery
\n\n", "votes": null, "replies": [] }, { "id": 161605, "author_name": "iglovikov", "author_url": "", "post_date": "02/14/2017 20:14:28", "content": "Semantic Segmentation of Earth Observation Data Using Multimodal and Multi-scale Deep Networks
", "votes": null, "replies": [] }, { "id": 161988, "author_name": "cubbeewan", "author_url": "", "post_date": "02/16/2017 18:06:11", "content": "Semantic Segmentation Of Aerial Images Wtih An Ensemble Of CNNs 2-head modified FCN + CRF
", "votes": null, "replies": [] }, { "id": 162489, "author_name": "velikolepno", "author_url": "", "post_date": "02/19/2017 13:00:52", "content": "http://juliandewit.github.io/kaggle-ndsb/ - some observations about using U-net for segmentation of biomedical images
", "votes": null, "replies": [] }, { "id": 163430, "author_name": "iglovikov", "author_url": "", "post_date": "02/24/2017 00:47:32", "content": "Better understanding of different bands.\n
https://www.mapbox.com/blog/putting-landsat-8-bands-to-work/
", "votes": null, "replies": [ { "id": 163565, "author_name": "ceperaang", "author_url": "", "post_date": "02/24/2017 15:23:05", "content": "Just to note: bands in the article (and picture above) are for Landsat-8 satellite, not for Worldview-3 we have here. And our bands are quite different.
", "votes": null, "replies": [] }, { "id": 163589, "author_name": "iglovikov", "author_url": "", "post_date": "02/24/2017 17:18:17", "content": "For sure, a numeration is quite different, but one may map Landsat-8 bands onto Worldview-3 using the wavelength.
", "votes": null, "replies": [] }, { "id": 164640, "author_name": "armamut", "author_url": "", "post_date": "03/01/2017 21:31:03", "content": "Here is a pdf file for Worldview-3 sensor wavebands.
\n\nhttps://www.spaceimagingme.com/downloads/sensors/datasheets/DG_WorldView3_DS_2014.pdf
", "votes": null, "replies": [] } ] }, { "id": 163821, "author_name": "cubbeewan", "author_url": "", "post_date": "02/26/2017 05:41:03", "content": "\n\nResults don't look very promising.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "159542": "This looks relevant:\n[Object Detection on SpaceNet][1]\n\n\n [1]: https://medium.com/the-downlinq/object-detection-on-spacenet-5e691961d257", "159919": "[Classification and Segmentation of Satellite Orthoimagery Using Convolutional Neural Networks][1] - looks like a nice review, discussing many design choices and how they affect performance. Check out citations too.\n\n[Scene Classification via a Gradient Boosting Random Convolutional Network Framework][2] - this is a slightly different problem - the task is just to classify which kind of scene is in a small patch. They also do gradient boosting of several networks, which might be applicable to other network architectures, maybe?\n\n\n [1]: http://www.mdpi.com/2072-4292/8/4/329/html\n [2]: https://www.researchgate.net/publication/283523609_Scene_Classification_via_a_Gradient_Boosting_Random_Convolutional_Network_Framework", "160170": "https://github.com/nshaud/DeepNetsForEO", "160270": "Facebook sharp mask paper referenced here https://code.facebook.com/posts/561187904071636/segmenting-and-refining-images-with-sharpmask/ also looks relevant. Interesting that their architecture is very similar to UNet.", "161035": "I really like this paper, some ideas like Atrous Lauer and CRF for post processing may potentially be applied to our problem.\n\n[DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs][1]\n\n\n [1]: https://arxiv.org/abs/1606.00915", "161087": "[A novel spectral index to automatically extract road networks from WorldView-2 satellite imagery][1]\n\n[Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery][2]\n\n [1]: http://www.sciencedirect.com/science/article/pii/S111098231400043X \"A novel spectral index to automatically extract road networks from WorldView-2 satellite imagery\"\n [2]: http://www.aari.ru/docs/pub/060804/xuh06.pdf \"Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery\"", "161605": "[Semantic Segmentation of Earth Observation Data Using Multimodal and Multi-scale Deep Networks][1]\n\n\n [1]: https://arxiv.org/abs/1609.06846", "161988": "[Semantic Segmentation Of Aerial Images Wtih An Ensemble Of CNNs][1] 2-head modified FCN + CRF\n\n\n [1]: http://intanto.net/publications/Marmanis_isprs16.pdf", "162489": "http://juliandewit.github.io/kaggle-ndsb/ - some observations about using U-net for segmentation of biomedical images", "163430": "Better understanding of different bands.\n![enter image description here][1]\n\nhttps://www.mapbox.com/blog/putting-landsat-8-bands-to-work/\n\n\n [1]: https://farm8.staticflickr.com/7309/9037121736_735a512d76_o.png", "163565": "Just to note: bands in the article (and picture above) are for Landsat-8 satellite, not for Worldview-3 we have here. And our bands are quite different.", "163589": "For sure, a numeration is quite different, but one may map Landsat-8 bands onto Worldview-3 using the wavelength.", "163821": "[Dropband: A Convolutional Neural Network With Data Augmentation For Scene Classification Of VHR Satellite Images][1] \n\nResults don't look very promising.\n [1]: http://proceedings.utwente.nl/403/1/Yang-DropBand-91.pdf", "164640": "Here is a pdf file for Worldview-3 sensor wavebands.\n\nhttps://www.spaceimagingme.com/downloads/sensors/datasheets/DG_WorldView3_DS_2014.pdf" }, "source": "meta" }