![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F069d738b7a7d9ede66fae61bda11ea61%2FTrain_data_Set01.PNG?generation=1576341900915374&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F782279656c959e0dc48b7ef9f7d78d82%2FTrain_Metadata.PNG?generation=1576341907637855&alt=media\")
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2234817%2F2e038e5b9017763c19025080a9fa834d%2FPublic_Test.PNG?generation=1576341910537769&alt=media\")
I have just downloaded the set01 of train data and the first video is aassnaulhq.mp4 which is also present in public test as first video .
\n\nThe JSON also says that its a FAKE video . Since the dataset is huge , how much can we ensure there is no leak ? Or Am I misunderstanding something about this competition ?
\n\nAnyone can clarify ?\n\n\n
I have just downloaded the set01 of train data and the first video is aassnaulhq.mp4 which is also present in public test as first video .
\n\nThe JSON also says that its a FAKE video . Since the dataset is huge , how much can we ensure there is no leak ? Or Am I misunderstanding something about this competition ?
\n\nAnyone can clarify ?\n\n\n
This appears to be true for the entire public validation set (i.e. the test_videos folder). All 400 test videos are present in the full training set.
(The notebook you submit isn't scored on these videos, so it's not really leakage.)
", "rawMarkdown": "This appears to be true for the entire public validation set (i.e. the `test_videos` folder). All 400 test videos are present in the full training set.\n\n(The notebook you submit isn't scored on these videos, so it's not really leakage.)", "votes": null }, { "id": "697301", "postDate": "12/17/2019 17:28:32", "content": "Hi! Thanks for your question. @humananalog is correct - the public test set is not included in scoring. It's just there to test your code against.
", "rawMarkdown": "Hi! Thanks for your question. @humananalog is correct - the public test set is not included in scoring. It's just there to test your code against.", "votes": null }, { "id": "1247550", "postDate": "03/21/2021 20:12:50", "content": "Hello, I am working on this competition now for testing a few things. I am a bit lost with the structure of the data, especially the test set and training set and I am not a native english speaker so just to be sure:
\n-the metadata.json file contains both the public training and the public testing set (aka videos in the training and testing set) ?
\n-the labels for the document in the testing set are also in the metadata.json file ? So the 400 documents cover both training set and testing set in this file ?
\nOr Am I getting it completely wrong and:
\n-the metadata.json is only for the videos in the training folder
\n-the submission.csv file is the one that actually contains the videos in the testing set along with their labels ?
Hello, I am working on this competition now for testing a few things. I am a bit lost with the structure of the data, especially the test set and training set and I am not a native english speaker so just to be sure:
\n-the metadata.json file contains both the public training and the public testing set (aka videos in the training and testing set) ?
\n-the labels for the document in the testing set are also in the metadata.json file ? So the 400 documents cover both training set and testing set in this file ?
\nOr Am I getting it completely wrong and:
\n-the metadata.json is only for the videos in the training folder
\n-the submission.csv file is the one that actually contains the videos in the testing set along with their labels ?
This appears to be true for the entire public validation set (i.e. the test_videos folder). All 400 test videos are present in the full training set.
(The notebook you submit isn't scored on these videos, so it's not really leakage.)
", "votes": null, "replies": [] }, { "id": 697301, "author_name": "philculliton", "author_url": "", "post_date": "12/17/2019 17:28:32", "content": "Hi! Thanks for your question. @humananalog is correct - the public test set is not included in scoring. It's just there to test your code against.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "695142": "I have just downloaded the set01 of train data and the first video is aassnaulhq.mp4 which is also present in public test as first video . \n\nThe JSON also says that its a FAKE video . Since the dataset is huge , how much can we ensure there is no leak ? Or Am I misunderstanding something about this competition ?\n\nAnyone can clarify ?\n\n\n", "695156": "This appears to be true for the entire public validation set (i.e. the `test_videos` folder). All 400 test videos are present in the full training set.\n\n(The notebook you submit isn't scored on these videos, so it's not really leakage.)", "697301": "Hi! Thanks for your question. @humananalog is correct - the public test set is not included in scoring. It's just there to test your code against.", "1247550": "Hello, I am working on this competition now for testing a few things. I am a bit lost with the structure of the data, especially the test set and training set and I am not a native english speaker so just to be sure:\n-the metadata.json file contains both the public training and the public testing set (aka videos in the training and testing set) ?\n-the labels for the document in the testing set are also in the metadata.json file ? So the 400 documents cover both training set and testing set in this file ?\nOr Am I getting it completely wrong and:\n-the metadata.json is only for the videos in the training folder\n-the submission.csv file is the one that actually contains the videos in the testing set along with their labels ?" }, "source": "meta" }