{ "id": 304686, "title": "Reduced Resolution Image Data (128 x 128, 256 x 256, 384 x 384) 🐋", "url": "/competitions/happy-whale-and-dolphin/discussion/304686", "author_name": "RDizzl3", "post_date": "2022-02-02T05:57:59.558000", "votes": 108, "comment_count": 23, "views": 0, "content": "

Hey all!

\n

I have created 3 different data sets for this competition. I have taken the original images and scaled them down to 3 different resolutions (128 x 128, 256 x 256 & 384 x 384). I wanted to provide these data sets to help anyone get started on their journey here. From previous competitions we have seen that higher resolutions generally have better model performance but these provide a great starting point.

\n

We can build our models starting at the 128 x 128 resolution. This can help iterate on ideas more quickly and also help us check our pipeline for bugs. Once we are confident our models are performing as expected at lower resolutions we can crank up the image sizes and continue experimenting. I suspect the resolutions for this competition might get fairly high (based on some of the image resolutions I saw in the original data set).

\n

These datasets are also a lot smaller in scale compared to the original dataset which should make it easier to download locally or even use in a Kaggle kernel. Please let me know if you have any feedback ✨

\n

Here are the dataset links:

\n
    \n
  1. (128 x 128 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-128x128
    2. \n
  2. (256 x 256 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-256x256
    3. \n
  3. (384 x 384 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-384x384
    4. \n
\n

Here is a very simple notebook as well that shows how to use the 128 x 128 image data. It loads in the image files and then shows that our image shape is (128, 128, 3) (which is what we want) for a single image and then plots the single image as well.

\n

https://www.kaggle.com/rdizzl3/happywhale-128-x-128-dataset-pilot

", "messages": [ { "id": 1672496, "postDate": "2022-02-02T05:57:59.560Z", "content": "

Hey all!

\n

I have created 3 different data sets for this competition. I have taken the original images and scaled them down to 3 different resolutions (128 x 128, 256 x 256 & 384 x 384). I wanted to provide these data sets to help anyone get started on their journey here. From previous competitions we have seen that higher resolutions generally have better model performance but these provide a great starting point.

\n

We can build our models starting at the 128 x 128 resolution. This can help iterate on ideas more quickly and also help us check our pipeline for bugs. Once we are confident our models are performing as expected at lower resolutions we can crank up the image sizes and continue experimenting. I suspect the resolutions for this competition might get fairly high (based on some of the image resolutions I saw in the original data set).

\n

These datasets are also a lot smaller in scale compared to the original dataset which should make it easier to download locally or even use in a Kaggle kernel. Please let me know if you have any feedback ✨

\n

Here are the dataset links:

\n
    \n
  1. (128 x 128 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-128x128
    2. \n
  2. (256 x 256 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-256x256
    3. \n
  3. (384 x 384 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-384x384
    4. \n
\n

Here is a very simple notebook as well that shows how to use the 128 x 128 image data. It loads in the image files and then shows that our image shape is (128, 128, 3) (which is what we want) for a single image and then plots the single image as well.

\n

https://www.kaggle.com/rdizzl3/happywhale-128-x-128-dataset-pilot

", "rawMarkdown": "Hey all!\n\nI have created 3 different data sets for this competition. I have taken the original images and scaled them down to 3 different resolutions `(128 x 128, 256 x 256 & 384 x 384)`. I wanted to provide these data sets to help anyone get started on their journey here. From previous competitions we have seen that higher resolutions generally have better model performance but these provide a great starting point. \n\nWe can build our models starting at the `128 x 128` resolution. This can help iterate on ideas more quickly and also help us check our pipeline for bugs. Once we are confident our models are performing as expected at lower resolutions we can crank up the image sizes and continue experimenting. I suspect the resolutions for this competition might get fairly high (based on some of the image resolutions I saw in the original data set).\n\nThese datasets are also a lot smaller in scale compared to the original dataset which should make it easier to download locally or even use in a Kaggle kernel. Please let me know if you have any feedback ✨\n\nHere are the dataset links: \n1. (128 x 128 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-128x128\n2. (256 x 256 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-256x256\n3. (384 x 384 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-384x384\n\nHere is a very simple notebook as well that shows how to use the 128 x 128 image data. It loads in the image files and then shows that our image shape is `(128, 128, 3)` (which is what we want) for a single image and then plots the single image as well.\n\nhttps://www.kaggle.com/rdizzl3/happywhale-128-x-128-dataset-pilot", "votes": 106 }, { "id": 1674593, "postDate": "2022-02-03T16:07:21.353Z", "content": "

@rdizzl3 it would be great if we could get 512512 and 768768 dataset or you could just share your nb. Would be off great help

", "rawMarkdown": "@rdizzl3 it would be great if we could get 512*512 and 768*768 dataset or you could just share your nb. Would be off great help ", "votes": 1 }, { "id": 1674481, "postDate": "2022-02-03T14:41:40.387Z", "content": "

Thanks!
\nAnd here are the original image sizes, for your convenience: https://www.kaggle.com/greendolphin/happywhale-2022-image-dims
\n@rdizzl3 feel free to add them to your dataset

", "rawMarkdown": "Thanks!\nAnd here are the original image sizes, for your convenience: https://www.kaggle.com/greendolphin/happywhale-2022-image-dims\n@rdizzl3 feel free to add them to your dataset", "votes": 1 }, { "id": 1672565, "postDate": "2022-02-02T06:59:55.023Z", "content": "

Thanks for the resized images datasets. The 3rd link (384 x 384 dataset) doesn't seem to work, can you please update that?

", "rawMarkdown": "Thanks for the resized images datasets. The 3rd link (384 x 384 dataset) doesn't seem to work, can you please update that?", "votes": 1, "replies": [ { "id": 1672580, "postDate": "2022-02-02T07:09:20.940Z", "content": "

@atharvaingle could you give it a try now? Could you check the other links as well?

", "rawMarkdown": "@atharvaingle could you give it a try now? Could you check the other links as well?", "votes": 2 }, { "id": 1672660, "postDate": "2022-02-02T08:19:49.987Z", "content": "

Thanks all links are working now!

", "rawMarkdown": "Thanks all links are working now!", "votes": 1 } ] }, { "id": 1672553, "postDate": "2022-02-02T06:53:01.807Z", "content": "

Thank you for sharing, you save our time!

", "rawMarkdown": "Thank you for sharing, you save our time! ", "votes": 1 }, { "id": 1685932, "postDate": "2022-02-11T16:22:20.290Z", "content": "

sharing my resize code here (preserving aspect ratio)
\nhttps://www.kaggle.com/ronigur/images-resizing

", "rawMarkdown": "sharing my resize code here (preserving aspect ratio)\nhttps://www.kaggle.com/ronigur/images-resizing", "votes": 2 }, { "id": 1673674, "postDate": "2022-02-02T21:50:53.200Z", "content": "

Thanks for sharing this. Quick question, which interpolation strategy was used to resize?

", "rawMarkdown": "Thanks for sharing this. Quick question, which interpolation strategy was used to resize? ", "votes": 2, "replies": [ { "id": 1673697, "postDate": "2022-02-02T22:15:34.533Z", "content": "

@ayuraj great question cv2.INTER_CUBIC

", "rawMarkdown": "@ayuraj great question `cv2.INTER_CUBIC`", "votes": 1 }, { "id": 1677005, "postDate": "2022-02-05T12:54:24.090Z", "content": "

Hey @RZizzl3 one more question - was the aspect ration was preserved while resizing?

\n

Apologies for nagging. :D

", "rawMarkdown": "Hey @RZizzl3 one more question - was the aspect ration was preserved while resizing? \n\nApologies for nagging. :D" } ] }, { "id": 1696985, "postDate": "2022-02-19T09:35:37.043Z", "content": "

Thanks for sharing. This will save a lot of time.

", "rawMarkdown": "Thanks for sharing. This will save a lot of time." }, { "id": 1695761, "postDate": "2022-02-18T10:49:40.263Z", "content": "

@rdizzl3 , first of all thanks for creating this dataset, could you please also share the image resize script for reference.

", "rawMarkdown": "@rdizzl3 , first of all thanks for creating this dataset, could you please also share the image resize script for reference." }, { "id": 1675380, "postDate": "2022-02-04T08:19:11.183Z", "content": "

Thank you for sharing these useful datasets! you save my time!

", "rawMarkdown": "Thank you for sharing these useful datasets! you save my time!" }, { "id": 1673961, "postDate": "2022-02-03T05:51:43.243Z", "content": "

Request to add train.csv file too

", "rawMarkdown": "Request to add train.csv file too ", "isDeleted": true }, { "id": 1679203, "postDate": "2022-02-07T04:39:38.880Z", "content": "

@rdizzl3 Thanks for sharing

", "rawMarkdown": "@rdizzl3 Thanks for sharing", "votes": 2 }, { "id": 1700863, "postDate": "2022-02-22T10:18:14.660Z", "content": "

thanks for it!
\nVery helpful!

", "rawMarkdown": "thanks for it! \nVery helpful!" }, { "id": 1693366, "postDate": "2022-02-16T16:18:10.067Z", "content": "

Thanks for sharing

", "rawMarkdown": "Thanks for sharing" }, { "id": 1685322, "postDate": "2022-02-11T07:39:48.687Z", "content": "

thank you for sharing

", "rawMarkdown": "thank you for sharing" }, { "id": 1679322, "postDate": "2022-02-07T06:47:08.823Z", "content": "

Thanks for sharing

", "rawMarkdown": "Thanks for sharing" }, { "id": 1678098, "postDate": "2022-02-06T08:54:31.103Z", "content": "

Thanks for sharing this with us.

", "rawMarkdown": "Thanks for sharing this with us." }, { "id": 1677882, "postDate": "2022-02-06T05:01:12.207Z", "content": "

Thanks for sharing

", "rawMarkdown": "Thanks for sharing" }, { "id": 1676863, "postDate": "2022-02-05T10:53:31.480Z", "content": "

Hi thank you for sharing the dataset

", "rawMarkdown": "Hi thank you for sharing the dataset" }, { "id": 1674465, "postDate": "2022-02-03T14:27:42.823Z", "content": "

Great work, thanks a lot

", "rawMarkdown": "Great work, thanks a lot" } ], "comments": [ { "id": 1674593, "author_name": "Mithil Salunkhe", "author_url": "", "post_date": "2022-02-03T16:07:21.353000", "content": "

@rdizzl3 it would be great if we could get 512512 and 768768 dataset or you could just share your nb. Would be off great help

", "votes": 1, "replies": [] }, { "id": 1674481, "author_name": "Marius Wanko", "author_url": "", "post_date": "2022-02-03T14:41:40.387000", "content": "

Thanks!
\nAnd here are the original image sizes, for your convenience: https://www.kaggle.com/greendolphin/happywhale-2022-image-dims
\n@rdizzl3 feel free to add them to your dataset

", "votes": 1, "replies": [] }, { "id": 1672565, "author_name": "Atharva Ingle", "author_url": "", "post_date": "2022-02-02T06:59:55.023000", "content": "

Thanks for the resized images datasets. The 3rd link (384 x 384 dataset) doesn't seem to work, can you please update that?

", "votes": 1, "replies": [ { "id": 1672580, "author_name": "RDizzl3", "author_url": "", "post_date": "2022-02-02T07:09:20.940000", "content": "

@atharvaingle could you give it a try now? Could you check the other links as well?

", "votes": 2, "replies": [] }, { "id": 1672660, "author_name": "Atharva Ingle", "author_url": "", "post_date": "2022-02-02T08:19:49.987000", "content": "

Thanks all links are working now!

", "votes": 1, "replies": [] } ] }, { "id": 1672553, "author_name": "Vadim Irtlach", "author_url": "", "post_date": "2022-02-02T06:53:01.807000", "content": "

Thank you for sharing, you save our time!

", "votes": 1, "replies": [] }, { "id": 1685932, "author_name": "Roni", "author_url": "", "post_date": "2022-02-11T16:22:20.290000", "content": "

sharing my resize code here (preserving aspect ratio)
\nhttps://www.kaggle.com/ronigur/images-resizing

", "votes": 2, "replies": [] }, { "id": 1673674, "author_name": "Ayush Thakur", "author_url": "", "post_date": "2022-02-02T21:50:53.200000", "content": "

Thanks for sharing this. Quick question, which interpolation strategy was used to resize?

", "votes": 2, "replies": [ { "id": 1673697, "author_name": "RDizzl3", "author_url": "", "post_date": "2022-02-02T22:15:34.533000", "content": "

@ayuraj great question cv2.INTER_CUBIC

", "votes": 1, "replies": [] }, { "id": 1677005, "author_name": "Ayush Thakur", "author_url": "", "post_date": "2022-02-05T12:54:24.090000", "content": "

Hey @RZizzl3 one more question - was the aspect ration was preserved while resizing?

\n

Apologies for nagging. :D

", "votes": 0, "replies": [] } ] }, { "id": 1696985, "author_name": "Shamim Ahamed", "author_url": "", "post_date": "2022-02-19T09:35:37.043000", "content": "

Thanks for sharing. This will save a lot of time.

", "votes": 0, "replies": [] }, { "id": 1695761, "author_name": "winmedals", "author_url": "", "post_date": "2022-02-18T10:49:40.263000", "content": "

@rdizzl3 , first of all thanks for creating this dataset, could you please also share the image resize script for reference.

", "votes": 0, "replies": [] }, { "id": 1675380, "author_name": "Ma Yuan", "author_url": "", "post_date": "2022-02-04T08:19:11.183000", "content": "

Thank you for sharing these useful datasets! you save my time!

", "votes": 0, "replies": [] }, { "id": 1673961, "author_name": "", "author_url": "", "post_date": "2022-02-03T05:51:43.243000", "content": "

Request to add train.csv file too

", "votes": 0, "replies": [] }, { "id": 1679203, "author_name": "Athar Sayed", "author_url": "", "post_date": "2022-02-07T04:39:38.880000", "content": "

@rdizzl3 Thanks for sharing

", "votes": 2, "replies": [] }, { "id": 1700863, "author_name": "Haru", "author_url": "", "post_date": "2022-02-22T10:18:14.660000", "content": "

thanks for it!
\nVery helpful!

", "votes": 0, "replies": [] }, { "id": 1693366, "author_name": "jzyztzn", "author_url": "", "post_date": "2022-02-16T16:18:10.067000", "content": "

Thanks for sharing

", "votes": 0, "replies": [] }, { "id": 1685322, "author_name": "Yuantao Yang", "author_url": "", "post_date": "2022-02-11T07:39:48.687000", "content": "

thank you for sharing

", "votes": 0, "replies": [] }, { "id": 1679322, "author_name": "Bithika Das", "author_url": "", "post_date": "2022-02-07T06:47:08.823000", "content": "

Thanks for sharing

", "votes": 0, "replies": [] }, { "id": 1678098, "author_name": "Ammar Abbasi1040", "author_url": "", "post_date": "2022-02-06T08:54:31.103000", "content": "

Thanks for sharing this with us.

", "votes": 0, "replies": [] }, { "id": 1677882, "author_name": "Abdul Manan", "author_url": "", "post_date": "2022-02-06T05:01:12.207000", "content": "

Thanks for sharing

", "votes": 0, "replies": [] }, { "id": 1676863, "author_name": "Ranpang", "author_url": "", "post_date": "2022-02-05T10:53:31.480000", "content": "

Hi thank you for sharing the dataset

", "votes": 0, "replies": [] }, { "id": 1674465, "author_name": "ChiangTyrol", "author_url": "", "post_date": "2022-02-03T14:27:42.823000", "content": "

Great work, thanks a lot

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1672496": "Hey all!\n\nI have created 3 different data sets for this competition. I have taken the original images and scaled them down to 3 different resolutions `(128 x 128, 256 x 256 & 384 x 384)`. I wanted to provide these data sets to help anyone get started on their journey here. From previous competitions we have seen that higher resolutions generally have better model performance but these provide a great starting point. \n\nWe can build our models starting at the `128 x 128` resolution. This can help iterate on ideas more quickly and also help us check our pipeline for bugs. Once we are confident our models are performing as expected at lower resolutions we can crank up the image sizes and continue experimenting. I suspect the resolutions for this competition might get fairly high (based on some of the image resolutions I saw in the original data set).\n\nThese datasets are also a lot smaller in scale compared to the original dataset which should make it easier to download locally or even use in a Kaggle kernel. Please let me know if you have any feedback ✨\n\nHere are the dataset links: \n1. (128 x 128 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-128x128\n2. (256 x 256 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-256x256\n3. (384 x 384 dataset) https://www.kaggle.com/rdizzl3/jpeg-happywhale-384x384\n\nHere is a very simple notebook as well that shows how to use the 128 x 128 image data. It loads in the image files and then shows that our image shape is `(128, 128, 3)` (which is what we want) for a single image and then plots the single image as well.\n\nhttps://www.kaggle.com/rdizzl3/happywhale-128-x-128-dataset-pilot", "1674593": "@rdizzl3 it would be great if we could get 512*512 and 768*768 dataset or you could just share your nb. Would be off great help ", "1674481": "Thanks!\nAnd here are the original image sizes, for your convenience: https://www.kaggle.com/greendolphin/happywhale-2022-image-dims\n@rdizzl3 feel free to add them to your dataset", "1672565": "Thanks for the resized images datasets. The 3rd link (384 x 384 dataset) doesn't seem to work, can you please update that?", "1672553": "Thank you for sharing, you save our time! ", "1685932": "sharing my resize code here (preserving aspect ratio)\nhttps://www.kaggle.com/ronigur/images-resizing", "1673674": "Thanks for sharing this. Quick question, which interpolation strategy was used to resize? ", "1696985": "Thanks for sharing. This will save a lot of time.", "1695761": "@rdizzl3 , first of all thanks for creating this dataset, could you please also share the image resize script for reference.", "1675380": "Thank you for sharing these useful datasets! you save my time!", "1673961": "Request to add train.csv file too ", "1679203": "@rdizzl3 Thanks for sharing", "1700863": "thanks for it! \nVery helpful!", "1693366": "Thanks for sharing", "1685322": "thank you for sharing", "1679322": "Thanks for sharing", "1678098": "Thanks for sharing this with us.", "1677882": "Thanks for sharing", "1676863": "Hi thank you for sharing the dataset", "1674465": "Great work, thanks a lot" } }