Welcome to the third edition of the Landmark Retrieval challenge! This year's competition is structured in a representation learning format: rather than creating a submission file with retrieved images, you will create a model that extracts a feature embedding for the images and submit the model via Kaggle Notebooks. Kaggle will run your model on a held-out test set, perform a k-nearest-neighbors lookup, and score the resulting embedding quality with mean average precision.
\nSince this is a new competition format, we expect there to be a break-in period and some unfamiliar details to understand. We will be releasing the evaluation metric code to assist with debugging, and the Landmark team will be releasing an example model baseline.
\nGood luck!
", "messages": [ { "id": 909884, "postDate": "2020-06-30T22:30:34.817Z", "content": "Welcome to the third edition of the Landmark Retrieval challenge! This year's competition is structured in a representation learning format: rather than creating a submission file with retrieved images, you will create a model that extracts a feature embedding for the images and submit the model via Kaggle Notebooks. Kaggle will run your model on a held-out test set, perform a k-nearest-neighbors lookup, and score the resulting embedding quality with mean average precision.
\nSince this is a new competition format, we expect there to be a break-in period and some unfamiliar details to understand. We will be releasing the evaluation metric code to assist with debugging, and the Landmark team will be releasing an example model baseline.
\nGood luck!
", "rawMarkdown": "Welcome to the third edition of the Landmark Retrieval challenge! This year's competition is structured in a representation learning format: rather than creating a submission file with retrieved images, you will create a model that extracts a feature embedding for the images and submit the model via Kaggle Notebooks. Kaggle will run your model on a held-out test set, perform a k-nearest-neighbors lookup, and score the resulting embedding quality with mean average precision.\n\nSince this is a new competition format, we expect there to be a break-in period and some unfamiliar details to understand. We will be releasing the evaluation metric code to assist with debugging, and the Landmark team will be releasing an example model baseline.\n\nGood luck!", "votes": 22 }, { "id": 910052, "postDate": "2020-07-01T01:35:31.550Z", "content": "Can't wait to see what everyone comes up with. Welcome, all!
", "rawMarkdown": "Can't wait to see what everyone comes up with. Welcome, all!", "votes": 5 }, { "id": 909904, "postDate": "2020-06-30T23:05:10.837Z", "content": "Welcome, everyone! We hope you will enjoy the challenge!
", "rawMarkdown": "Welcome, everyone! We hope you will enjoy the challenge!", "votes": 6 }, { "id": 909898, "postDate": "2020-06-30T22:55:32.627Z", "content": "welcome, everyone! Have fun with the competition :)
", "rawMarkdown": "welcome, everyone! Have fun with the competition :)", "votes": 6 }, { "id": 911415, "postDate": "2020-07-01T18:04:03.170Z", "content": "\n\n\nWhen can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)
\n
Check out our posts Your First Submission as a basic starting point and Your Second Submission for end to end instructions on training your own model.
", "rawMarkdown": "> When can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)\n\nCheck out our posts [Your First Submission](https://www.kaggle.com/c/landmark-retrieval-2020/discussion/163341) as a basic starting point and [Your Second Submission](https://www.kaggle.com/c/landmark-retrieval-2020/discussion/163350) for end to end instructions on training your own model.", "votes": 3, "replies": [ { "id": 911425, "postDate": "2020-07-01T18:13:21.117Z", "content": "Cool, thanks! :)
", "rawMarkdown": "Cool, thanks! :)\n" } ] }, { "id": 910323, "postDate": "2020-07-01T05:22:48.240Z", "content": "Welcome, everyone! Have fun! Let us know if you have any further questions!
", "rawMarkdown": "Welcome, everyone! Have fun! Let us know if you have any further questions!\n\n", "votes": 3 }, { "id": 918203, "postDate": "2020-07-07T04:37:58.993Z", "content": "Excited for this challenge.
", "rawMarkdown": "Excited for this challenge.", "votes": 1 }, { "id": 910300, "postDate": "2020-07-01T05:06:51.770Z", "content": "Looking forward for this awesome challenge
", "rawMarkdown": "Looking forward for this awesome challenge", "votes": 1 }, { "id": 910473, "postDate": "2020-07-01T06:58:51.903Z", "content": "Thanks for the competition, but I think restricting model submissions to Tensorflow is not a good choice, considering that there is ONNX for interoperability.
\n\nThe statement (https://www.kaggle.com/c/landmark-retrieval-2020/data):
\n\n> Your model must be named submission.zip and be compatible with TensorFlow 2.2. The submission.zip should contain all files and directories created by the tf.saved_model_save function using Tensorflow's SavedModel format.
", "rawMarkdown": "Thanks for the competition, but I think restricting model submissions to Tensorflow is not a good choice, considering that there is ONNX for interoperability. \n\nThe statement (https://www.kaggle.com/c/landmark-retrieval-2020/data):\n\n> Your model must be named submission.zip and be compatible with TensorFlow 2.2. The submission.zip should contain all files and directories created by the tf.saved_model_save function using Tensorflow's SavedModel format.", "votes": 1 }, { "id": 911911, "postDate": "2020-07-02T05:23:56.917Z", "content": "The big G has come back. Looking forward for sexiest collections of codes.
", "rawMarkdown": "The big G has come back. Looking forward for sexiest collections of codes." }, { "id": 911693, "postDate": "2020-07-02T00:12:38.947Z", "content": "Google comes again with awesome surprises.
", "rawMarkdown": "Google comes again with awesome surprises. " }, { "id": 911135, "postDate": "2020-07-01T15:30:30.960Z", "content": "Hello Big fellas !! welcome
", "rawMarkdown": "Hello Big fellas !! welcome " }, { "id": 911071, "postDate": "2020-07-01T14:52:42.143Z", "content": "When can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)
", "rawMarkdown": "When can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)" }, { "id": 916135, "postDate": "2020-07-05T11:57:21.743Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 910052, "author_name": "Cam Askew", "author_url": "", "post_date": "2020-07-01T01:35:31.550000", "content": "Can't wait to see what everyone comes up with. Welcome, all!
", "votes": 5, "replies": [] }, { "id": 909904, "author_name": "Tobias Weyand", "author_url": "", "post_date": "2020-06-30T23:05:10.837000", "content": "Welcome, everyone! We hope you will enjoy the challenge!
", "votes": 6, "replies": [] }, { "id": 909898, "author_name": "Andre Araujo", "author_url": "", "post_date": "2020-06-30T22:55:32.627000", "content": "welcome, everyone! Have fun with the competition :)
", "votes": 6, "replies": [] }, { "id": 911415, "author_name": "Tobias Weyand", "author_url": "", "post_date": "2020-07-01T18:04:03.170000", "content": "\n\n\nWhen can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)
\n
Check out our posts Your First Submission as a basic starting point and Your Second Submission for end to end instructions on training your own model.
", "votes": 3, "replies": [ { "id": 911425, "author_name": "Gajendra Saraswat", "author_url": "", "post_date": "2020-07-01T18:13:21.117000", "content": "Cool, thanks! :)
", "votes": 0, "replies": [] } ] }, { "id": 910323, "author_name": "Bingyi Cao", "author_url": "", "post_date": "2020-07-01T05:22:48.240000", "content": "Welcome, everyone! Have fun! Let us know if you have any further questions!
", "votes": 3, "replies": [] }, { "id": 918203, "author_name": "Divyanshu Jhawar", "author_url": "", "post_date": "2020-07-07T04:37:58.993000", "content": "Excited for this challenge.
", "votes": 1, "replies": [] }, { "id": 910300, "author_name": "shivyshiv", "author_url": "", "post_date": "2020-07-01T05:06:51.770000", "content": "Looking forward for this awesome challenge
", "votes": 1, "replies": [] }, { "id": 910473, "author_name": "Aadhav Vignesh", "author_url": "", "post_date": "2020-07-01T06:58:51.903000", "content": "Thanks for the competition, but I think restricting model submissions to Tensorflow is not a good choice, considering that there is ONNX for interoperability.
\n\nThe statement (https://www.kaggle.com/c/landmark-retrieval-2020/data):
\n\n> Your model must be named submission.zip and be compatible with TensorFlow 2.2. The submission.zip should contain all files and directories created by the tf.saved_model_save function using Tensorflow's SavedModel format.
", "votes": 1, "replies": [] }, { "id": 911911, "author_name": "TNBS92", "author_url": "", "post_date": "2020-07-02T05:23:56.917000", "content": "The big G has come back. Looking forward for sexiest collections of codes.
", "votes": 0, "replies": [] }, { "id": 911693, "author_name": "", "author_url": "", "post_date": "2020-07-02T00:12:38.947000", "content": "Google comes again with awesome surprises.
", "votes": 0, "replies": [] }, { "id": 911135, "author_name": "Kanchan Sarkar", "author_url": "", "post_date": "2020-07-01T15:30:30.960000", "content": "Hello Big fellas !! welcome
", "votes": 0, "replies": [] }, { "id": 911071, "author_name": "Gajendra Saraswat", "author_url": "", "post_date": "2020-07-01T14:52:42.143000", "content": "When can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)
", "votes": 0, "replies": [] }, { "id": 916135, "author_name": "", "author_url": "", "post_date": "2020-07-05T11:57:21.743000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "909884": "Welcome to the third edition of the Landmark Retrieval challenge! This year's competition is structured in a representation learning format: rather than creating a submission file with retrieved images, you will create a model that extracts a feature embedding for the images and submit the model via Kaggle Notebooks. Kaggle will run your model on a held-out test set, perform a k-nearest-neighbors lookup, and score the resulting embedding quality with mean average precision.\n\nSince this is a new competition format, we expect there to be a break-in period and some unfamiliar details to understand. We will be releasing the evaluation metric code to assist with debugging, and the Landmark team will be releasing an example model baseline.\n\nGood luck!", "910052": "Can't wait to see what everyone comes up with. Welcome, all!", "909904": "Welcome, everyone! We hope you will enjoy the challenge!", "909898": "welcome, everyone! Have fun with the competition :)", "911415": "> When can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)\n\nCheck out our posts [Your First Submission](https://www.kaggle.com/c/landmark-retrieval-2020/discussion/163341) as a basic starting point and [Your Second Submission](https://www.kaggle.com/c/landmark-retrieval-2020/discussion/163350) for end to end instructions on training your own model.", "910323": "Welcome, everyone! Have fun! Let us know if you have any further questions!\n\n", "918203": "Excited for this challenge.", "910300": "Looking forward for this awesome challenge", "910473": "Thanks for the competition, but I think restricting model submissions to Tensorflow is not a good choice, considering that there is ONNX for interoperability. \n\nThe statement (https://www.kaggle.com/c/landmark-retrieval-2020/data):\n\n> Your model must be named submission.zip and be compatible with TensorFlow 2.2. The submission.zip should contain all files and directories created by the tf.saved_model_save function using Tensorflow's SavedModel format.", "911911": "The big G has come back. Looking forward for sexiest collections of codes.", "911693": "Google comes again with awesome surprises. ", "911135": "Hello Big fellas !! welcome ", "911071": "When can we get a baseline starter notebook? I understand it takes time but it would be great to have it asap as I feel it would be great to start working for this amazing and different competition! :)", "916135": "" } }