We all know the biggest problem with this competition is the training data. It's Whack and hard to train on. So I have taken the coordinates and remade the data set with all their classes and images. Each Image has a folder with the lions from that image and a CSV with its' correct classes.
\n\nHere is the file, via MediaFire\nIf you need it from elsewhere you can contact me.
\n\nAgain:
\nhttp://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz
Reference is also Here
", "messages": [ { "id": 180969, "postDate": "2017-05-08T03:08:27.610Z", "content": "We all know the biggest problem with this competition is the training data. It's Whack and hard to train on. So I have taken the coordinates and remade the data set with all their classes and images. Each Image has a folder with the lions from that image and a CSV with its' correct classes.
\n\nHere is the file, via MediaFire\nIf you need it from elsewhere you can contact me.
\n\nAgain:
\nhttp://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz
Reference is also Here
", "rawMarkdown": "We all know the biggest problem with this competition is the training data. It's Whack and hard to train on. So I have taken the coordinates and remade the data set with all their classes and images. Each Image has a folder with the lions from that image and a CSV with its' correct classes.\n\n[Here][1] is the file, via MediaFire\nIf you need it from elsewhere you can contact me.\n\nAgain: \n[http://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz][2]\n\nReference is also [Here][3]\n\n [1]: http://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz\n [2]: http://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz\n [3]: https://www.kaggle.com/kingburrito666/noaa-pipelined-data", "votes": 6 }, { "id": 181175, "postDate": "2017-05-08T18:12:13.493Z", "content": "That's awesome! I might actually invest some time (mine an my computers' ) into this competition after all. Thanks for sharing, that's very generous of you.
", "rawMarkdown": "That's awesome! I might actually invest some time (mine an my computers' ) into this competition after all. Thanks for sharing, that's very generous of you. ", "votes": 1, "replies": [ { "id": 181189, "postDate": "2017-05-08T18:59:20.017Z", "content": "Thanks, Bojan! I'm currently working on a Deeplearning4J model right now. DL4J, because Java is my home language and it can utilize different GPU architectures.
", "rawMarkdown": "Thanks, Bojan! I'm currently working on a Deeplearning4J model right now. [DL4J][1], because Java is my home language and it can utilize different GPU architectures.\n\n\n [1]: https://deeplearning4j.org/index.html", "votes": 1 }, { "id": 181229, "postDate": "2017-05-08T21:36:34.317Z", "content": "That's interesting. Java is on my looong list of languages to learn, but haven't gotten around it yet.
\n\nFrom what I understand, TensorFlow was designed specifically with diversity and heterogeneity of GPU architectures and setups in mind, so you might want to take a look at it at some point. I primarily work with Keras, as I like its easy user interface, and there are plenty of pretrained models for it.
", "rawMarkdown": "That's interesting. Java is on my looong list of languages to learn, but haven't gotten around it yet. \n\nFrom what I understand, TensorFlow was designed specifically with diversity and heterogeneity of GPU architectures and setups in mind, so you might want to take a look at it at some point. I primarily work with Keras, as I like its easy user interface, and there are plenty of pretrained models for it. ", "votes": 1 }, { "id": 325971, "postDate": "2018-05-09T06:16:14.730Z", "content": "Hey Liam, did you succeed in implementing a dl4j model for this competition? If so, is it accessible to the public?
", "rawMarkdown": "Hey Liam, did you succeed in implementing a dl4j model for this competition? If so, is it accessible to the public?" } ] }, { "id": 189147, "postDate": "2017-06-05T11:09:28.140Z", "content": "Hey. What did you use to create the dataset of images? Do you have some codes etc. ?
\n\nMy proxy blocks access to mediafire but the extracted images will be very useful for training the models.
", "rawMarkdown": "Hey. What did you use to create the dataset of images? Do you have some codes etc. ? \n\nMy proxy blocks access to mediafire but the extracted images will be very useful for training the models. " }, { "id": 188503, "postDate": "2017-06-03T02:23:16.623Z", "content": "Hi, thanks so much for your sharing. But I have a little problem. In each folder, there are images \"0.jpg\" and \"z_0.jpg\". Are all the files beginning with \"z\" negative samples? Thank you.
", "rawMarkdown": "Hi, thanks so much for your sharing. But I have a little problem. In each folder, there are images \"0.jpg\" and \"z_0.jpg\". Are all the files beginning with \"z\" negative samples? Thank you.", "replies": [ { "id": 188511, "postDate": "2017-06-03T03:45:48.023Z", "content": "anything with a Z means it's negative and is NOT in the CSV
", "rawMarkdown": "anything with a Z means it's negative and is NOT in the CSV" } ] }, { "id": 180989, "postDate": "2017-05-08T06:15:44.480Z", "rawMarkdown": "", "votes": 2, "isDeleted": true }, { "id": 181539, "postDate": "2017-05-09T22:22:42.780Z", "content": "Hi! thank you a lot! Just awesome!
", "rawMarkdown": "Hi! thank you a lot! Just awesome!\n" } ], "comments": [ { "id": 181175, "author_name": "Bojan Tunguz", "author_url": "", "post_date": "2017-05-08T18:12:13.493000", "content": "That's awesome! I might actually invest some time (mine an my computers' ) into this competition after all. Thanks for sharing, that's very generous of you.
", "votes": 1, "replies": [ { "id": 181189, "author_name": "Liam Larsen", "author_url": "", "post_date": "2017-05-08T18:59:20.017000", "content": "Thanks, Bojan! I'm currently working on a Deeplearning4J model right now. DL4J, because Java is my home language and it can utilize different GPU architectures.
", "votes": 1, "replies": [] }, { "id": 181229, "author_name": "Bojan Tunguz", "author_url": "", "post_date": "2017-05-08T21:36:34.317000", "content": "That's interesting. Java is on my looong list of languages to learn, but haven't gotten around it yet.
\n\nFrom what I understand, TensorFlow was designed specifically with diversity and heterogeneity of GPU architectures and setups in mind, so you might want to take a look at it at some point. I primarily work with Keras, as I like its easy user interface, and there are plenty of pretrained models for it.
", "votes": 1, "replies": [] }, { "id": 325971, "author_name": "holgerbrandl", "author_url": "", "post_date": "2018-05-09T06:16:14.730000", "content": "Hey Liam, did you succeed in implementing a dl4j model for this competition? If so, is it accessible to the public?
", "votes": 0, "replies": [] } ] }, { "id": 189147, "author_name": "saurabh bansal", "author_url": "", "post_date": "2017-06-05T11:09:28.140000", "content": "Hey. What did you use to create the dataset of images? Do you have some codes etc. ?
\n\nMy proxy blocks access to mediafire but the extracted images will be very useful for training the models.
", "votes": 0, "replies": [] }, { "id": 188503, "author_name": "Aurora", "author_url": "", "post_date": "2017-06-03T02:23:16.623000", "content": "Hi, thanks so much for your sharing. But I have a little problem. In each folder, there are images \"0.jpg\" and \"z_0.jpg\". Are all the files beginning with \"z\" negative samples? Thank you.
", "votes": 0, "replies": [ { "id": 188511, "author_name": "Liam Larsen", "author_url": "", "post_date": "2017-06-03T03:45:48.023000", "content": "anything with a Z means it's negative and is NOT in the CSV
", "votes": 0, "replies": [] } ] }, { "id": 180989, "author_name": "", "author_url": "", "post_date": "2017-05-08T06:15:44.480000", "content": "", "votes": 2, "replies": [] }, { "id": 181539, "author_name": "MatveySafroshkin", "author_url": "", "post_date": "2017-05-09T22:22:42.780000", "content": "Hi! thank you a lot! Just awesome!
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "180969": "We all know the biggest problem with this competition is the training data. It's Whack and hard to train on. So I have taken the coordinates and remade the data set with all their classes and images. Each Image has a folder with the lions from that image and a CSV with its' correct classes.\n\n[Here][1] is the file, via MediaFire\nIf you need it from elsewhere you can contact me.\n\nAgain: \n[http://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz][2]\n\nReference is also [Here][3]\n\n [1]: http://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz\n [2]: http://www.mediafire.com/file/qb2p0z94az91cgp/Train_Data.tar.gz\n [3]: https://www.kaggle.com/kingburrito666/noaa-pipelined-data", "181175": "That's awesome! I might actually invest some time (mine an my computers' ) into this competition after all. Thanks for sharing, that's very generous of you. ", "189147": "Hey. What did you use to create the dataset of images? Do you have some codes etc. ? \n\nMy proxy blocks access to mediafire but the extracted images will be very useful for training the models. ", "188503": "Hi, thanks so much for your sharing. But I have a little problem. In each folder, there are images \"0.jpg\" and \"z_0.jpg\". Are all the files beginning with \"z\" negative samples? Thank you.", "180989": "", "181539": "Hi! thank you a lot! Just awesome!\n" } }