{ "id": 223477, "title": "Resized and reshaped datasets - 128/256px, JPG/PNG, padded/original ratio", "url": "/competitions/bms-molecular-translation/discussion/223477", "author_name": "", "post_date": "2021-03-04T03:21:38.795934100Z", "votes": 29, "comment_count": 6, "views": 0, "content": "

I've resized both all the images (training and test) into JPG images using variants of this notebook.

\n

By using the datasets below, you are accepting the rules of this competition!

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Size (Pixels)FormatShapeDataset
64JPGSquare (padded)Link
128JPGOriginal ratioLink
128JPGSquare (padded)Link
128JPGSquare (stretched)Link
128PNGSquare (padded)Link
224JPGSquare (padded)Link
256JPGOriginal ratioLink
256JPGSquare (padded)Link
\n

I'm planning to publish datasets in various formats (original ratio, stretched, padded), and sizes (128px, 256px, and perhaps higher). I'm running into various problems when I try to save the files so it's taking longer than expected to process them. When I successfully resolve them I will update this post with links to the respective datasets.

\n
\n

Note: The datasets marked with a star * might still be processing; if it's shown as unavailable please come back in a few hours!

\n
", "messages": [ { "id": "1225889", "postDate": "03/04/2021 03:21:38", "content": "

I've resized both all the images (training and test) into JPG images using variants of this notebook.

\n

By using the datasets below, you are accepting the rules of this competition!

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Size (Pixels)FormatShapeDataset
64JPGSquare (padded)Link
128JPGOriginal ratioLink
128JPGSquare (padded)Link
128JPGSquare (stretched)Link
128PNGSquare (padded)Link
224JPGSquare (padded)Link
256JPGOriginal ratioLink
256JPGSquare (padded)Link
\n

I'm planning to publish datasets in various formats (original ratio, stretched, padded), and sizes (128px, 256px, and perhaps higher). I'm running into various problems when I try to save the files so it's taking longer than expected to process them. When I successfully resolve them I will update this post with links to the respective datasets.

\n
\n

Note: The datasets marked with a star * might still be processing; if it's shown as unavailable please come back in a few hours!

\n
", "rawMarkdown": "I've resized both all the images (training and test) into JPG images using variants of [this notebook](https://www.kaggle.com/xhlulu/bms-molecular-resizing-and-reshaping). \n\nBy using the datasets below, **you are accepting the rules of this competition!**\n\n| Size (Pixels) | Format | Shape | Dataset |\n|---------------|--------|--------------------|------------------------------------------------------------------------------|\n| 64 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-164px-jpg-padded) |\n| 128 | JPG | Original ratio | [Link](https://www.kaggle.com/xhlulu/bms-molecular-128px-jpg-original-ratio) |\n| 128 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-128px-jpg-padded) |\n| 128 | JPG | Square (stretched) | [Link](https://www.kaggle.com/xhlulu/bms-128-jpg-stretch) |\n| 128 | PNG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-128px-png-padded) |\n| 224 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-224px-jpg-padded) |\n| 256 | JPG | Original ratio | [Link](https://www.kaggle.com/xhlulu/bms-molecular-256px-jpg-original-ratio) |\n| 256 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-256px-jpg-padded) |\n\nI'm planning to publish datasets in various formats (original ratio, stretched, padded), and sizes (128px, 256px, and perhaps higher). I'm running into various problems when I try to save the files so it's taking longer than expected to process them. When I successfully resolve them I will update this post with links to the respective datasets.\n\n\n> Note: The datasets marked with a star `*` might still be processing; if it's shown as unavailable please come back in a few hours!", "votes": null }, { "id": "1226073", "postDate": "03/04/2021 07:59:18", "content": "

Can you share your associated code for doing it?
\nThanks

", "rawMarkdown": "Can you share your associated code for doing it? \nThanks", "votes": null }, { "id": "1226077", "postDate": "03/04/2021 08:01:22", "content": "

Thanks a lot @xhlulu , It will be a great help.
\nI was trying to do something similar… Good that you have done it already. Also what is the ETA you are experiencing for the same?
\nBecause I was also looking to generate TFRecords for this…

", "rawMarkdown": "Thanks a lot @xhlulu , It will be a great help.\nI was trying to do something similar... Good that you have done it already. Also what is the ETA you are experiencing for the same?\nBecause I was also looking to generate TFRecords for this...", "votes": null }, { "id": "1227561", "postDate": "03/05/2021 16:45:21", "content": "

Hi, I updated the post with a link to the notebook :)

", "rawMarkdown": "Hi, I updated the post with a link to the notebook :)", "votes": null }, { "id": "1227563", "postDate": "03/05/2021 16:45:44", "content": "

On a multi-core setup it takes around 4h to run the notebook (and some more time to process afterward)

", "rawMarkdown": "On a multi-core setup it takes around 4h to run the notebook (and some more time to process afterward)", "votes": null }, { "id": "1227715", "postDate": "03/05/2021 19:35:27", "content": "

Maybe a white-padded version could work better since the black padding can give the model some clues and it could end up cheating by looking at it instead of looking at the molecules.

", "rawMarkdown": "Maybe a white-padded version could work better since the black padding can give the model some clues and it could end up cheating by looking at it instead of looking at the molecules.", "votes": null }, { "id": "1227909", "postDate": "03/05/2021 23:39:14", "content": "

Yeah I wanted to do a white padding, but didn't input the correct color parameter… Now i'm rerunning but it's probably gonna take a while!

", "rawMarkdown": "Yeah I wanted to do a white padding, but didn't input the correct `color` parameter... Now i'm rerunning but it's probably gonna take a while!", "votes": null } ], "comments": [ { "id": 1226073, "author_name": "mrinath", "author_url": "", "post_date": "03/04/2021 07:59:18", "content": "

Can you share your associated code for doing it?
\nThanks

", "votes": null, "replies": [ { "id": 1227561, "author_name": "xhlulu", "author_url": "", "post_date": "03/05/2021 16:45:21", "content": "

Hi, I updated the post with a link to the notebook :)

", "votes": null, "replies": [] } ] }, { "id": 1226077, "author_name": "manabendrarout", "author_url": "", "post_date": "03/04/2021 08:01:22", "content": "

Thanks a lot @xhlulu , It will be a great help.
\nI was trying to do something similar… Good that you have done it already. Also what is the ETA you are experiencing for the same?
\nBecause I was also looking to generate TFRecords for this…

", "votes": null, "replies": [ { "id": 1227563, "author_name": "xhlulu", "author_url": "", "post_date": "03/05/2021 16:45:44", "content": "

On a multi-core setup it takes around 4h to run the notebook (and some more time to process afterward)

", "votes": null, "replies": [] } ] }, { "id": 1227715, "author_name": "pabloppp", "author_url": "", "post_date": "03/05/2021 19:35:27", "content": "

Maybe a white-padded version could work better since the black padding can give the model some clues and it could end up cheating by looking at it instead of looking at the molecules.

", "votes": null, "replies": [ { "id": 1227909, "author_name": "xhlulu", "author_url": "", "post_date": "03/05/2021 23:39:14", "content": "

Yeah I wanted to do a white padding, but didn't input the correct color parameter… Now i'm rerunning but it's probably gonna take a while!

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1225889": "I've resized both all the images (training and test) into JPG images using variants of [this notebook](https://www.kaggle.com/xhlulu/bms-molecular-resizing-and-reshaping). \n\nBy using the datasets below, **you are accepting the rules of this competition!**\n\n| Size (Pixels) | Format | Shape | Dataset |\n|---------------|--------|--------------------|------------------------------------------------------------------------------|\n| 64 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-164px-jpg-padded) |\n| 128 | JPG | Original ratio | [Link](https://www.kaggle.com/xhlulu/bms-molecular-128px-jpg-original-ratio) |\n| 128 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-128px-jpg-padded) |\n| 128 | JPG | Square (stretched) | [Link](https://www.kaggle.com/xhlulu/bms-128-jpg-stretch) |\n| 128 | PNG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-128px-png-padded) |\n| 224 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-224px-jpg-padded) |\n| 256 | JPG | Original ratio | [Link](https://www.kaggle.com/xhlulu/bms-molecular-256px-jpg-original-ratio) |\n| 256 | JPG | Square (padded) | [Link](https://www.kaggle.com/xhlulu/bms-molecular-256px-jpg-padded) |\n\nI'm planning to publish datasets in various formats (original ratio, stretched, padded), and sizes (128px, 256px, and perhaps higher). I'm running into various problems when I try to save the files so it's taking longer than expected to process them. When I successfully resolve them I will update this post with links to the respective datasets.\n\n\n> Note: The datasets marked with a star `*` might still be processing; if it's shown as unavailable please come back in a few hours!", "1226073": "Can you share your associated code for doing it? \nThanks", "1226077": "Thanks a lot @xhlulu , It will be a great help.\nI was trying to do something similar... Good that you have done it already. Also what is the ETA you are experiencing for the same?\nBecause I was also looking to generate TFRecords for this...", "1227561": "Hi, I updated the post with a link to the notebook :)", "1227563": "On a multi-core setup it takes around 4h to run the notebook (and some more time to process afterward)", "1227715": "Maybe a white-padded version could work better since the black padding can give the model some clues and it could end up cheating by looking at it instead of looking at the molecules.", "1227909": "Yeah I wanted to do a white padding, but didn't input the correct `color` parameter... Now i'm rerunning but it's probably gonna take a while!" }, "source": "meta" }