I downloaded the Oxford 102 category Flower dataset and converted it into TFRecords for everyone in the Kaggle dataset here.
\n\nI was concerned that there is some overlap with this dataset and the competition's data but Kaggle has said here that we can use this dataset. Since this is a powerful external dataset, it is only fair to make it available to everyone.
\n\nEnjoy!!
\n\n=============
\n\nUPDATE: I explain how to add external training datasets to your notebook here
", "messages": [ { "id": 796818, "postDate": "2020-04-04T00:28:06.100Z", "content": "I downloaded the Oxford 102 category Flower dataset and converted it into TFRecords for everyone in the Kaggle dataset here.
\n\nI was concerned that there is some overlap with this dataset and the competition's data but Kaggle has said here that we can use this dataset. Since this is a powerful external dataset, it is only fair to make it available to everyone.
\n\nEnjoy!!
\n\n=============
\n\nUPDATE: I explain how to add external training datasets to your notebook here
", "rawMarkdown": "I downloaded the Oxford 102 category Flower dataset and converted it into TFRecords for everyone in the Kaggle dataset [here][1]. \n\nI was concerned that there is some overlap with this dataset and the competition's data but Kaggle has said [here][2] that we can use this dataset. Since this is a powerful external dataset, it is only fair to make it available to everyone.\n\nEnjoy!!\n\n=============\n\nUPDATE: I explain how to add external training datasets to your notebook [here][3]\n\n[1]: https://www.kaggle.com/cdeotte/oxford-flowers-tfrecords\n[2]: https://www.kaggle.com/c/flower-classification-with-tpus/discussion/134701#769297\n[3]: https://www.kaggle.com/c/flower-classification-with-tpus/discussion/148329", "votes": 37 }, { "id": 799797, "postDate": "2020-04-06T19:01:43.240Z", "content": "FYI, the Oxford 102 flowers dataset is almost entirely contained in this 104 flowers dataset.
", "rawMarkdown": "FYI, the Oxford 102 flowers dataset is almost entirely contained in this 104 flowers dataset.", "votes": 3, "replies": [ { "id": 801490, "postDate": "2020-04-08T14:41:45.207Z", "content": "Ohh thanks for the information, in the training data only or testing data also?
", "rawMarkdown": "Ohh thanks for the information, in the training data only or testing data also?\n" }, { "id": 811264, "postDate": "2020-04-17T18:05:39.583Z", "content": "Lol! Do you expect a reply to that?
\n\nLet's hope it has images of testing data as well
", "rawMarkdown": "Lol! Do you expect a reply to that?\n\nLet's hope it has images of testing data as well" }, { "id": 812859, "postDate": "2020-04-19T06:58:44.323Z", "content": "Why would you hope it to include testing data? (leak)
", "rawMarkdown": "Why would you hope it to include testing data? (leak)" }, { "id": 813707, "postDate": "2020-04-19T22:46:29.873Z", "content": "I would assume it's because predicting a seen image will theoretically produce the corresponding (seen) label?
", "rawMarkdown": "I would assume it's because predicting a seen image will theoretically produce the corresponding (seen) label? " }, { "id": 813886, "postDate": "2020-04-20T05:24:16.097Z", "content": "I need to try this dataset then, I guess. Otherwise, If It contains test dataset, then whoever does not use it is in a disadvantage.
", "rawMarkdown": "I need to try this dataset then, I guess. Otherwise, If It contains test dataset, then whoever does not use it is in a disadvantage. " }, { "id": 814521, "postDate": "2020-04-20T18:15:02.083Z", "content": "Yeah my thinking is if you use include this dataset then it would be the equivalent of training model on training data then also predicting on that same data. so let's say this new dataset contains the majority of the images in the test set, you could potentially overfit on purpose and damn near win the challenge lol
", "rawMarkdown": "Yeah my thinking is if you use include this dataset then it would be the equivalent of training model on training data then also predicting on that same data. so let's say this new dataset contains the majority of the images in the test set, you could potentially overfit on purpose and damn near win the challenge lol" }, { "id": 816207, "postDate": "2020-04-22T07:16:24.073Z", "content": "or just hash the images, compare and predict with 100% accuracy? :P
", "rawMarkdown": "or just hash the images, compare and predict with 100% accuracy? :P " }, { "id": 816211, "postDate": "2020-04-22T07:21:22.723Z", "content": "says the person 15th in this competition 🤔 there may be something here...
", "rawMarkdown": "says the person 15th in this competition 🤔 there may be something here... " } ] }, { "id": 832199, "postDate": "2020-05-04T00:38:19.750Z", "content": "UPDATE: I explain how to add external training datasets to your notebook here
", "rawMarkdown": "UPDATE: I explain how to add external training datasets to your notebook [here][1]\n\n[1]: https://www.kaggle.com/c/flower-classification-with-tpus/discussion/148329", "votes": 1 }, { "id": 825873, "postDate": "2020-04-29T09:48:37.883Z", "content": "hey guys i have a doubt. I am not able to download the labels for the flowers. i have downloaded the images but the labels are a .mat file and its a shortcut too but just a 1kb file. How do i get the labels?
", "rawMarkdown": "hey guys i have a doubt. I am not able to download the labels for the flowers. i have downloaded the images but the labels are a .mat file and its a shortcut too but just a 1kb file. How do i get the labels?", "votes": 1 }, { "id": 796947, "postDate": "2020-04-04T05:33:14.233Z", "content": "Thank you @cdeotte ..This is very helpful
", "rawMarkdown": "Thank you @cdeotte ..This is very helpful ", "votes": 1 }, { "id": 815616, "postDate": "2020-04-21T17:16:32.463Z", "content": "A starter notebook made available for anyone who wants to use this external dataset.
", "rawMarkdown": "A [starter notebook](https://www.kaggle.com/kurianbenoy/training-with-oxford-external-data-keras-efnet/) made available for anyone who wants to use this external dataset." }, { "id": 813096, "postDate": "2020-04-19T11:25:38.333Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 799797, "author_name": "Martin Görner", "author_url": "", "post_date": "2020-04-06T19:01:43.240000", "content": "FYI, the Oxford 102 flowers dataset is almost entirely contained in this 104 flowers dataset.
", "votes": 3, "replies": [ { "id": 801490, "author_name": "Kamal Chhirang", "author_url": "", "post_date": "2020-04-08T14:41:45.207000", "content": "Ohh thanks for the information, in the training data only or testing data also?
", "votes": 0, "replies": [] }, { "id": 811264, "author_name": "Kurian Benoy", "author_url": "", "post_date": "2020-04-17T18:05:39.583000", "content": "Lol! Do you expect a reply to that?
\n\nLet's hope it has images of testing data as well
", "votes": 0, "replies": [] }, { "id": 812859, "author_name": "Kamal Chhirang", "author_url": "", "post_date": "2020-04-19T06:58:44.323000", "content": "Why would you hope it to include testing data? (leak)
", "votes": 0, "replies": [] }, { "id": 813707, "author_name": "Rashan Arshad", "author_url": "", "post_date": "2020-04-19T22:46:29.873000", "content": "I would assume it's because predicting a seen image will theoretically produce the corresponding (seen) label?
", "votes": 0, "replies": [] }, { "id": 813886, "author_name": "Kamal Chhirang", "author_url": "", "post_date": "2020-04-20T05:24:16.097000", "content": "I need to try this dataset then, I guess. Otherwise, If It contains test dataset, then whoever does not use it is in a disadvantage.
", "votes": 0, "replies": [] }, { "id": 814521, "author_name": "Rashan Arshad", "author_url": "", "post_date": "2020-04-20T18:15:02.083000", "content": "Yeah my thinking is if you use include this dataset then it would be the equivalent of training model on training data then also predicting on that same data. so let's say this new dataset contains the majority of the images in the test set, you could potentially overfit on purpose and damn near win the challenge lol
", "votes": 0, "replies": [] }, { "id": 816207, "author_name": "Kamal Chhirang", "author_url": "", "post_date": "2020-04-22T07:16:24.073000", "content": "or just hash the images, compare and predict with 100% accuracy? :P
", "votes": 0, "replies": [] }, { "id": 816211, "author_name": "Rashan Arshad", "author_url": "", "post_date": "2020-04-22T07:21:22.723000", "content": "says the person 15th in this competition 🤔 there may be something here...
", "votes": 0, "replies": [] } ] }, { "id": 832199, "author_name": "Chris Deotte", "author_url": "", "post_date": "2020-05-04T00:38:19.750000", "content": "UPDATE: I explain how to add external training datasets to your notebook here
", "votes": 1, "replies": [] }, { "id": 825873, "author_name": "Sumanth", "author_url": "", "post_date": "2020-04-29T09:48:37.883000", "content": "hey guys i have a doubt. I am not able to download the labels for the flowers. i have downloaded the images but the labels are a .mat file and its a shortcut too but just a 1kb file. How do i get the labels?
", "votes": 1, "replies": [] }, { "id": 796947, "author_name": "GSD", "author_url": "", "post_date": "2020-04-04T05:33:14.233000", "content": "Thank you @cdeotte ..This is very helpful
", "votes": 1, "replies": [] }, { "id": 815616, "author_name": "Kurian Benoy", "author_url": "", "post_date": "2020-04-21T17:16:32.463000", "content": "A starter notebook made available for anyone who wants to use this external dataset.
", "votes": 0, "replies": [] }, { "id": 813096, "author_name": "", "author_url": "", "post_date": "2020-04-19T11:25:38.333000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "796818": "I downloaded the Oxford 102 category Flower dataset and converted it into TFRecords for everyone in the Kaggle dataset [here][1]. \n\nI was concerned that there is some overlap with this dataset and the competition's data but Kaggle has said [here][2] that we can use this dataset. Since this is a powerful external dataset, it is only fair to make it available to everyone.\n\nEnjoy!!\n\n=============\n\nUPDATE: I explain how to add external training datasets to your notebook [here][3]\n\n[1]: https://www.kaggle.com/cdeotte/oxford-flowers-tfrecords\n[2]: https://www.kaggle.com/c/flower-classification-with-tpus/discussion/134701#769297\n[3]: https://www.kaggle.com/c/flower-classification-with-tpus/discussion/148329", "799797": "FYI, the Oxford 102 flowers dataset is almost entirely contained in this 104 flowers dataset.", "832199": "UPDATE: I explain how to add external training datasets to your notebook [here][1]\n\n[1]: https://www.kaggle.com/c/flower-classification-with-tpus/discussion/148329", "825873": "hey guys i have a doubt. I am not able to download the labels for the flowers. i have downloaded the images but the labels are a .mat file and its a shortcut too but just a 1kb file. How do i get the labels?", "796947": "Thank you @cdeotte ..This is very helpful ", "815616": "A [starter notebook](https://www.kaggle.com/kurianbenoy/training-with-oxford-external-data-keras-efnet/) made available for anyone who wants to use this external dataset.", "813096": "" } }