Hi,
\nI am new to kaggle and, perhaps, my question is obvious, but I don't understand the entire process of submission.
\nFirst of all, what do I submit - notebook or results?
\nThis is a code competition and, as I could understand, my code will be run automatically on entire dataset by kaggle, without my participation. Is that correct? In that case, I have to provide the notebook that could be run as a script, not cell by cell. Do I get it right?
\nI guess that there are detailed answers and instructions, which resolve all my questions. I'll very appreciate if anybody can provide me with links.
Hi,
\nI am new to kaggle and, perhaps, my question is obvious, but I don't understand the entire process of submission.
\nFirst of all, what do I submit - notebook or results?
\nThis is a code competition and, as I could understand, my code will be run automatically on entire dataset by kaggle, without my participation. Is that correct? In that case, I have to provide the notebook that could be run as a script, not cell by cell. Do I get it right?
\nI guess that there are detailed answers and instructions, which resolve all my questions. I'll very appreciate if anybody can provide me with links.
Kaggle Course page is a place I would recommend you go through
\nGood luck
", "rawMarkdown": "yurikreinin \n\n[Kaggle Course](https://www.kaggle.com/learn) page is a place I would recommend you go through\n- Python course: walks you through how to do a submission with the \n- Intro to ML an Intermediate ML walks you through the steps of typically done in a competition.\n\nGood luck", "votes": null }, { "id": "1647982", "postDate": "01/13/2022 03:20:37", "content": "You submit a notebook which outputs the predictions. In this competition, there is a tiny bit of special API code you must include in your notebook to properly make a submission.
\nYou will save your notebook however you wish to code it. Then on the competition submission page, you'll chose the notebook and click the submit button. The submission and scoring process will execute in the background.
\nKeep in mind that this is an internet off competition, so you must toggle the internet off in your notebook.
", "rawMarkdown": "You submit a notebook which outputs the predictions. In this competition, there is a tiny bit of special API code you must include in your notebook to properly make a submission. \n\nYou will save your notebook however you wish to code it. Then on the competition submission page, you'll chose the notebook and click the submit button. The submission and scoring process will execute in the background.\n\nKeep in mind that this is an *internet off* competition, so you must toggle the internet off in your notebook.", "votes": null }, { "id": "1648001", "postDate": "01/13/2022 03:46:15", "content": "David,
\nThank you for your reply.
\nI understand that I have to interact with API and I can see the example. But I don't get very simple thing: my code is notebook containing different cells, one which is derived from example. But there are others. For instance, I load my pre-trained model and I do some pre or post-processing etc. If I submit script (module) and not notebook - it would be transparent: functions, classes, global variables and then prediction sequence derived from the example under main, so it can be run in background. But how it is expected to work with notebook and not a script?
You don't need an entry point when you're using the notebook editor in Notebook mode, like you would in a normal script. If your notebook runs without error, submit it the way it is. There is no need to change it.
", "rawMarkdown": "You don't need an entry point when you're using the notebook editor in Notebook mode, like you would in a normal script. If your notebook runs without error, submit it the way it is. There is no need to change it.", "votes": null }, { "id": "1648054", "postDate": "01/13/2022 04:55:53", "content": "Thank you, David.
\nIf I got you right, it means that my entire notebook is running like it is a script from top to bottom.
Hi I followed all the steps but I keep getting a submission scoring error. Can someone advice please @davidbroberts
", "rawMarkdown": "Hi I followed all the steps but I keep getting a submission scoring error. Can someone advice please @davidbroberts", "votes": null }, { "id": "1666641", "postDate": "01/28/2022 00:31:37", "content": "@ulricharmel,
\nThere are some things that you should double-check.
Because testing doesn't contain any bounding boxes, I would recommend to double-check the annotations string that you create, write the code that:
\nThanks for the advices.
\nBut why are you sure of points 2 and 3 (:.2f and :.0f format)? I seem to be able to use float numbers without limiting decimal places (have score after submission).
I've received submission error until I changed coordinates to be integer.
\nRegarding to confidence, it may contain more than 2 decimal digits after floating point, I just provided the :.2f format as a recommendation.
Weird. I've tried without rounding and with rounding of bounding boxes. It worked this way and that. In the second case (for integer bounding boxes), the score was 0.001 higher.
", "rawMarkdown": "Weird. I've tried without rounding and with rounding of bounding boxes. It worked this way and that. In the second case (for integer bounding boxes), the score was 0.001 higher.", "votes": null } ], "comments": [ { "id": 1647865, "author_name": "yvonnef", "author_url": "", "post_date": "01/13/2022 01:29:00", "content": "\nKaggle Course page is a place I would recommend you go through
\nGood luck
", "votes": null, "replies": [] }, { "id": 1647982, "author_name": "davidbroberts", "author_url": "", "post_date": "01/13/2022 03:20:37", "content": "You submit a notebook which outputs the predictions. In this competition, there is a tiny bit of special API code you must include in your notebook to properly make a submission.
\nYou will save your notebook however you wish to code it. Then on the competition submission page, you'll chose the notebook and click the submit button. The submission and scoring process will execute in the background.
\nKeep in mind that this is an internet off competition, so you must toggle the internet off in your notebook.
", "votes": null, "replies": [ { "id": 1648001, "author_name": "yurikreinin", "author_url": "", "post_date": "01/13/2022 03:46:15", "content": "David,
\nThank you for your reply.
\nI understand that I have to interact with API and I can see the example. But I don't get very simple thing: my code is notebook containing different cells, one which is derived from example. But there are others. For instance, I load my pre-trained model and I do some pre or post-processing etc. If I submit script (module) and not notebook - it would be transparent: functions, classes, global variables and then prediction sequence derived from the example under main, so it can be run in background. But how it is expected to work with notebook and not a script?
You don't need an entry point when you're using the notebook editor in Notebook mode, like you would in a normal script. If your notebook runs without error, submit it the way it is. There is no need to change it.
", "votes": null, "replies": [] }, { "id": 1648054, "author_name": "yurikreinin", "author_url": "", "post_date": "01/13/2022 04:55:53", "content": "Thank you, David.
\nIf I got you right, it means that my entire notebook is running like it is a script from top to bottom.
Hi I followed all the steps but I keep getting a submission scoring error. Can someone advice please @davidbroberts
", "votes": null, "replies": [] }, { "id": 1666641, "author_name": "yurikreinin", "author_url": "", "post_date": "01/28/2022 00:31:37", "content": "@ulricharmel,
\nThere are some things that you should double-check.
Because testing doesn't contain any bounding boxes, I would recommend to double-check the annotations string that you create, write the code that:
\nThanks for the advices.
\nBut why are you sure of points 2 and 3 (:.2f and :.0f format)? I seem to be able to use float numbers without limiting decimal places (have score after submission).
I've received submission error until I changed coordinates to be integer.
\nRegarding to confidence, it may contain more than 2 decimal digits after floating point, I just provided the :.2f format as a recommendation.
Weird. I've tried without rounding and with rounding of bounding boxes. It worked this way and that. In the second case (for integer bounding boxes), the score was 0.001 higher.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1647733": "Hi,\nI am new to kaggle and, perhaps, my question is obvious, but I don't understand the entire process of submission.\nFirst of all, what do I submit - notebook or results?\nThis is a code competition and, as I could understand, my code will be run automatically on entire dataset by kaggle, without my participation. Is that correct? In that case, I have to provide the notebook that could be run as a script, not cell by cell. Do I get it right?\nI guess that there are detailed answers and instructions, which resolve all my questions. I'll very appreciate if anybody can provide me with links.", "1647865": "yurikreinin \n\n[Kaggle Course](https://www.kaggle.com/learn) page is a place I would recommend you go through\n- Python course: walks you through how to do a submission with the \n- Intro to ML an Intermediate ML walks you through the steps of typically done in a competition.\n\nGood luck", "1647982": "You submit a notebook which outputs the predictions. In this competition, there is a tiny bit of special API code you must include in your notebook to properly make a submission. \n\nYou will save your notebook however you wish to code it. Then on the competition submission page, you'll chose the notebook and click the submit button. The submission and scoring process will execute in the background.\n\nKeep in mind that this is an *internet off* competition, so you must toggle the internet off in your notebook.", "1648001": "David,\nThank you for your reply.\nI understand that I have to interact with API and I can see the example. But I don't get very simple thing: my code is notebook containing different cells, one which is derived from example. But there are others. For instance, I load my pre-trained model and I do some pre or post-processing etc. If I submit script (module) and not notebook - it would be transparent: functions, classes, global variables and then prediction sequence derived from the example under __main__, so it can be run in background. But how it is expected to work with notebook and not a script?", "1648050": "You don't need an entry point when you're using the notebook editor in Notebook mode, like you would in a normal script. If your notebook runs without error, submit it the way it is. There is no need to change it.", "1648054": "Thank you, David.\nIf I got you right, it means that my entire notebook is running like it is a script from top to bottom.", "1666556": "Hi I followed all the steps but I keep getting a submission scoring error. Can someone advice please @davidbroberts", "1666641": "ulricharmel,\nThere are some things that you should double-check.\n1. All values are \"space\" separated. No commas or anything else\n2. The first value for each bounding box (confidence) should be in floating point format (**:.2f**)\n3. The other 4 values (x y width height) has to be integers. No floating point is allowed. So, if your prediction returns floating point value for the rectangle - use **:.0f**. Actually, you can use this format for integer type as well, so just use **:.0f** format.\n\nBecause testing doesn't contain any bounding boxes, I would recommend to double-check the annotations string that you create, write the code that:\n1. Opens some file from dataset, which contains a few bounding boxes\n2. Run your prediction procedure\n3. Create the annotations string with exactly the same code, as in submission\n4. Print and review it", "1668815": "Thanks for the advices.\nBut why are you sure of points 2 and 3 (:.2f and :.0f format)? I seem to be able to use float numbers without limiting decimal places (have score after submission).", "1668884": "I've received ***submission erro***r until I changed coordinates to be integer.\nRegarding to *confidence*, it may contain more than 2 decimal digits after floating point, I just provided the **:.2f** format as a recommendation.", "1668985": "Weird. I've tried without rounding and with rounding of bounding boxes. It worked this way and that. In the second case (for integer bounding boxes), the score was 0.001 higher." }, "source": "meta" }