Resized(512x512) images to download if anyone likes to run/modify/tune @mrgloom's model: \nhttps://drive.google.com/open?id=0B15wedujp1cQVGR6Z1RQX2t5Nms
\n\nPlease take a look at the sample resized images in test_images_512x512_sample folder before download. They may not clear enough for fancy CNN architectures.
", "messages": [ { "id": "191717", "postDate": "06/11/2017 15:20:04", "content": "Resized(512x512) images to download if anyone likes to run/modify/tune @mrgloom's model: \nhttps://drive.google.com/open?id=0B15wedujp1cQVGR6Z1RQX2t5Nms
\n\nPlease take a look at the sample resized images in test_images_512x512_sample folder before download. They may not clear enough for fancy CNN architectures.
", "rawMarkdown": "Resized(512x512) images to download if anyone likes to run/modify/tune [@mrgloom's model][1]: \nhttps://drive.google.com/open?id=0B15wedujp1cQVGR6Z1RQX2t5Nms\n\nPlease take a look at the sample resized images in test_images_512x512_sample folder before download. They may not clear enough for fancy CNN architectures.\n\n [1]: https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count/discussion/33900", "votes": null }, { "id": "191732", "postDate": "06/11/2017 16:01:56", "content": "It's cool，but i havn't machine at hand，I want to have another silver medal. What a pity.
", "rawMarkdown": "It's cool，but i havn't machine at hand，I want to have another silver medal. What a pity.", "votes": null }, { "id": "191733", "postDate": "06/11/2017 16:06:22", "content": "What a pity +1
", "rawMarkdown": "What a pity +1", "votes": null }, { "id": "192008", "postDate": "06/12/2017 13:19:51", "content": "Congratulations on trying to save this competition. It is unfortunate that managers have not thought of this.
", "rawMarkdown": "Congratulations on trying to save this competition. It is unfortunate that managers have not thought of this.", "votes": null }, { "id": "192012", "postDate": "06/12/2017 13:23:18", "content": ":) I cannot find 200 Gb space on my PC so decided to resize them on AWS and download. I hope this can help people in the same situation. However, the resized images are not clear enough.
", "rawMarkdown": ":) I cannot find 200 Gb space on my PC so decided to resize them on AWS and download. I hope this can help people in the same situation. However, the resized images are not clear enough.", "votes": null }, { "id": "192018", "postDate": "06/12/2017 13:31:03", "content": "Hi MechCoder, there is a better way. Instead of resizing the whole image if you cut only the patches containing at least one dot your dataset will reduce dramatically. I can share my script if think that will help.
", "rawMarkdown": "Hi MechCoder, there is a better way. Instead of resizing the whole image if you cut only the patches containing at least one dot your dataset will reduce dramatically. I can share my script if think that will help.", "votes": null }, { "id": "192025", "postDate": "06/12/2017 13:40:46", "content": "amaia, thanks! can you please share the script?
", "rawMarkdown": "amaia, thanks! can you please share the script?", "votes": null }, { "id": "192031", "postDate": "06/12/2017 13:53:44", "content": "@amaia, that's sounds amazing. Would love to read your script. I am new to computer vision and willing to learn everything.
", "rawMarkdown": "amaia, that's sounds amazing. Would love to read your script. I am new to computer vision and willing to learn everything.", "votes": null }, { "id": "192035", "postDate": "06/12/2017 14:16:47", "content": "Here it goes. This is an extract from the version I used in the beginning to obtain bounding boxes from training set. You'll need to adjust paths and also lines 68-69 with the exact format you want your updated dot coordinates (relative to the patch coordinates). The coords.csv file is supplied in the same repository, it is the same shared in the beginning of the competition.
\n\nhttps://github.com/cxz/noaa/blob/master/train_patches.py\nhttps://github.com/cxz/noaa/blob/master/data/coords.csv
", "rawMarkdown": "Here it goes. This is an extract from the version I used in the beginning to obtain bounding boxes from training set. You'll need to adjust paths and also lines 68-69 with the exact format you want your updated dot coordinates (relative to the patch coordinates). The coords.csv file is supplied in the same repository, it is the same shared in the beginning of the competition.\n\nhttps://github.com/cxz/noaa/blob/master/train_patches.py\nhttps://github.com/cxz/noaa/blob/master/data/coords.csv", "votes": null }, { "id": "192038", "postDate": "06/12/2017 14:20:50", "content": "@MechCoder, not much computer vision, just a loop checking whether the patch contains dots or not (on the assumption that there are large areas in training set not useful for training).
", "rawMarkdown": "MechCoder, not much computer vision, just a loop checking whether the patch contains dots or not (on the assumption that there are large areas in training set not useful for training).", "votes": null }, { "id": "192044", "postDate": "06/12/2017 14:30:45", "content": "Thanks for sharing, @amaia!
", "rawMarkdown": "Thanks for sharing, @amaia!", "votes": null }, { "id": "192950", "postDate": "06/15/2017 07:10:08", "content": "Thanks, now it's easier for me to have a try )
", "rawMarkdown": "Thanks, now it's easier for me to have a try )", "votes": null }, { "id": "192983", "postDate": "06/15/2017 09:22:57", "content": "Enjoy!
", "rawMarkdown": "Enjoy!", "votes": null }, { "id": "193116", "postDate": "06/15/2017 16:48:15", "content": "@amaia, could you share some insights about dealing with testing data?
", "rawMarkdown": "amaia, could you share some insights about dealing with testing data?", "votes": null } ], "comments": [ { "id": 191732, "author_name": "sfx7rail", "author_url": "", "post_date": "06/11/2017 16:01:56", "content": "It's cool，but i havn't machine at hand，I want to have another silver medal. What a pity.
", "votes": null, "replies": [ { "id": 191733, "author_name": "shujian", "author_url": "", "post_date": "06/11/2017 16:06:22", "content": "What a pity +1
", "votes": null, "replies": [] } ] }, { "id": 192008, "author_name": "paulorzp", "author_url": "", "post_date": "06/12/2017 13:19:51", "content": "Congratulations on trying to save this competition. It is unfortunate that managers have not thought of this.
", "votes": null, "replies": [ { "id": 192012, "author_name": "shujian", "author_url": "", "post_date": "06/12/2017 13:23:18", "content": ":) I cannot find 200 Gb space on my PC so decided to resize them on AWS and download. I hope this can help people in the same situation. However, the resized images are not clear enough.
", "votes": null, "replies": [] }, { "id": 192018, "author_name": "aamaia", "author_url": "", "post_date": "06/12/2017 13:31:03", "content": "Hi MechCoder, there is a better way. Instead of resizing the whole image if you cut only the patches containing at least one dot your dataset will reduce dramatically. I can share my script if think that will help.
", "votes": null, "replies": [] }, { "id": 192025, "author_name": "cjansen", "author_url": "", "post_date": "06/12/2017 13:40:46", "content": "amaia, thanks! can you please share the script?
", "votes": null, "replies": [] }, { "id": 192031, "author_name": "shujian", "author_url": "", "post_date": "06/12/2017 13:53:44", "content": "@amaia, that's sounds amazing. Would love to read your script. I am new to computer vision and willing to learn everything.
", "votes": null, "replies": [] }, { "id": 192035, "author_name": "aamaia", "author_url": "", "post_date": "06/12/2017 14:16:47", "content": "Here it goes. This is an extract from the version I used in the beginning to obtain bounding boxes from training set. You'll need to adjust paths and also lines 68-69 with the exact format you want your updated dot coordinates (relative to the patch coordinates). The coords.csv file is supplied in the same repository, it is the same shared in the beginning of the competition.
\n\nhttps://github.com/cxz/noaa/blob/master/train_patches.py\nhttps://github.com/cxz/noaa/blob/master/data/coords.csv
", "votes": null, "replies": [] }, { "id": 192038, "author_name": "aamaia", "author_url": "", "post_date": "06/12/2017 14:20:50", "content": "@MechCoder, not much computer vision, just a loop checking whether the patch contains dots or not (on the assumption that there are large areas in training set not useful for training).
", "votes": null, "replies": [] }, { "id": 192044, "author_name": "shujian", "author_url": "", "post_date": "06/12/2017 14:30:45", "content": "Thanks for sharing, @amaia!
", "votes": null, "replies": [] }, { "id": 193116, "author_name": "shujian", "author_url": "", "post_date": "06/15/2017 16:48:15", "content": "@amaia, could you share some insights about dealing with testing data?
", "votes": null, "replies": [] } ] }, { "id": 192950, "author_name": "infinitewing", "author_url": "", "post_date": "06/15/2017 07:10:08", "content": "Thanks, now it's easier for me to have a try )
", "votes": null, "replies": [ { "id": 192983, "author_name": "shujian", "author_url": "", "post_date": "06/15/2017 09:22:57", "content": "Enjoy!
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "191717": "Resized(512x512) images to download if anyone likes to run/modify/tune [@mrgloom's model][1]: \nhttps://drive.google.com/open?id=0B15wedujp1cQVGR6Z1RQX2t5Nms\n\nPlease take a look at the sample resized images in test_images_512x512_sample folder before download. They may not clear enough for fancy CNN architectures.\n\n [1]: https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count/discussion/33900", "191732": "It's cool，but i havn't machine at hand，I want to have another silver medal. What a pity.", "191733": "What a pity +1", "192008": "Congratulations on trying to save this competition. It is unfortunate that managers have not thought of this.", "192012": ":) I cannot find 200 Gb space on my PC so decided to resize them on AWS and download. I hope this can help people in the same situation. However, the resized images are not clear enough.", "192018": "Hi MechCoder, there is a better way. Instead of resizing the whole image if you cut only the patches containing at least one dot your dataset will reduce dramatically. I can share my script if think that will help.", "192025": "amaia, thanks! can you please share the script?", "192031": "amaia, that's sounds amazing. Would love to read your script. I am new to computer vision and willing to learn everything.", "192035": "Here it goes. This is an extract from the version I used in the beginning to obtain bounding boxes from training set. You'll need to adjust paths and also lines 68-69 with the exact format you want your updated dot coordinates (relative to the patch coordinates). The coords.csv file is supplied in the same repository, it is the same shared in the beginning of the competition.\n\nhttps://github.com/cxz/noaa/blob/master/train_patches.py\nhttps://github.com/cxz/noaa/blob/master/data/coords.csv", "192038": "MechCoder, not much computer vision, just a loop checking whether the patch contains dots or not (on the assumption that there are large areas in training set not useful for training).", "192044": "Thanks for sharing, @amaia!", "192950": "Thanks, now it's easier for me to have a try )", "192983": "Enjoy!", "193116": "amaia, could you share some insights about dealing with testing data?" }, "source": "meta" }