{ "id": 239918, "title": "Resized images in JPG/PNG and different sizes", "url": "/competitions/siim-covid19-detection/discussion/239918", "author_name": "xhlulu", "post_date": "2021-05-18T03:52:37.249000", "votes": 125, "comment_count": 27, "views": 0, "content": "

I resized the images in this competition to JPG and PNG files in different sizes. Before resizing, I apply VOI LUT (see Raddar's notebook) whenever the information is available (which doesn't seem to be always the case), as well as monochrome fixing (see previous notebook).

\n

Click here for my preprocessing notebook

\n

The notebook can be easily modified to generate different sizes, aspect ratios and image format. Below is a list of datasets that were created by (variants of) this notebook:

\n", "messages": [ { "id": 1312430, "postDate": "2021-05-18T03:52:37.250Z", "content": "

I resized the images in this competition to JPG and PNG files in different sizes. Before resizing, I apply VOI LUT (see Raddar's notebook) whenever the information is available (which doesn't seem to be always the case), as well as monochrome fixing (see previous notebook).

\n

Click here for my preprocessing notebook

\n

The notebook can be easily modified to generate different sizes, aspect ratios and image format. Below is a list of datasets that were created by (variants of) this notebook:

\n", "rawMarkdown": "I resized the images in this competition to JPG and PNG files in different sizes. Before resizing, I apply VOI LUT (see [Raddar's notebook](https://www.kaggle.com/raddar/convert-dicom-to-np-array-the-correct-way)) whenever the information is available (which doesn't seem to be always the case), as well as monochrome fixing (see previous notebook).\n\n[Click here for my preprocessing notebook](https://www.kaggle.com/xhlulu/siim-covid-19-convert-to-jpg)\n\nThe notebook can be easily modified to generate different sizes, aspect ratios and image format. Below is a list of datasets that were created by (variants of) this notebook:\n* [SIIM COVID-19: Resized to 256px JPG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-256px-jpg)\n* [SIIM COVID-19: Resized to 512px JPG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-512px-jpg)\n* [SIIM COVID-19: Resized to 1024px JPG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-1024px-jpg)\n* [SIIM COVID-19: Resized to 224px PNG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-224px-png)\n* [SIIM COVID-19: Resized to 256px PNG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-256px-png)\n* [SIIM COVID-19: Resized to 512px PNG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-512px-png)", "votes": 125 }, { "id": 1314809, "postDate": "2021-05-19T11:49:07.257Z", "content": "

The rumour is @xhlulu is working part-time for Kaggle now 😜

\n

This should be great for kagglers like me who are low on bandwidth and GPU all the time ;)

", "rawMarkdown": "The rumour is @xhlulu is working part-time for Kaggle now 😜\n\nThis should be great for kagglers like me who are low on bandwidth and GPU all the time ;) ", "votes": 7 }, { "id": 1326630, "postDate": "2021-05-28T15:22:32.160Z", "content": "

Hey @xhlulu , thanks for making the dataset public. Just one concern, I went through your preprocessing notebook and it seems that keep_raito=False.

\n

I am getting weird location for bounding boxes. Here, have a look
\n\"Image\"

\n

IMO, the original images are pretty big. So, while reducing the size, aspect ratio & rounding errors will quickly add up.
\nDo let me know what you think? Is anyone else facing the same issue? If no, then its quite possible that I might be doing something wrong πŸ˜…

", "rawMarkdown": "Hey @xhlulu , thanks for making the dataset public. Just one concern, I went through your preprocessing notebook and it seems that `keep_raito=False`. \n\nI am getting weird location for bounding boxes. Here, have a look\n![Image](https://i.imgur.com/TXOkH4y.png)\n\nIMO, the original images are pretty big. So, while reducing the size, aspect ratio & rounding errors will quickly add up. \nDo let me know what you think? Is anyone else facing the same issue? If no, then its quite possible that I might be doing something wrong πŸ˜…\n", "votes": 5, "replies": [ { "id": 1360774, "postDate": "2021-06-22T10:42:16.630Z", "content": "

I haven't faced such an issue so far.

", "rawMarkdown": "I haven't faced such an issue so far. " } ] }, { "id": 1313116, "postDate": "2021-05-18T12:12:54.590Z", "content": "

Hey @xhlulu,
\nReally appreciate you creating the script for converting dicom image data πŸ˜„

\n

I've created a dataset of augmented images based on the jpg files you created!

\n

Augmentation types : RandomSunFlare, RandomFog, RandomBrightness, RandomCrop, Rotate, RGBShift, RandomSnow, HorizontalFlip, VerticalFlip, RandomContrast, HSV

", "rawMarkdown": "Hey @xhlulu,\nReally appreciate you creating the script for converting dicom image data πŸ˜„\n\nI've created a dataset of [augmented images](https://www.kaggle.com/ruchi798/siimfisabiorsna-covid19-detection-augmented) based on the jpg files you created! \n\n**Augmentation types** : RandomSunFlare, RandomFog, RandomBrightness, RandomCrop, Rotate, RGBShift, RandomSnow, HorizontalFlip, VerticalFlip, RandomContrast, HSV", "votes": 4, "replies": [ { "id": 1313299, "postDate": "2021-05-18T13:54:10.860Z", "content": "

Thanks for sharing. :)
\nAugmented images has also 256 image size?

", "rawMarkdown": "Thanks for sharing. :)\n`Augmented images` has also `256` image size?", "votes": 1 }, { "id": 1313329, "postDate": "2021-05-18T14:10:11.800Z", "content": "

Yes @piantic πŸ˜„

", "rawMarkdown": "Yes @piantic πŸ˜„", "votes": 1 }, { "id": 1313531, "postDate": "2021-05-18T15:44:17.670Z", "content": "

Cool, thanks for sharing! I think it would be nice if you can share the notebook too so people can try out different configurations of augmentations :)

", "rawMarkdown": "Cool, thanks for sharing! I think it would be nice if you can share the notebook too so people can try out different configurations of augmentations :)", "votes": 1 }, { "id": 1313566, "postDate": "2021-05-18T15:57:01.303Z", "content": "

I've included that segment in my notebook.
\nHope this helps 😌

", "rawMarkdown": "I've included that segment in my [notebook](https://www.kaggle.com/ruchi798/siim-covid-19-detection-eda-data-augmentation).\nHope this helps 😌 ", "votes": 3 } ] }, { "id": 1316242, "postDate": "2021-05-20T11:51:07.243Z", "content": "

@xhlulu, the link to 256x256 PNG is broken. It says 404 not found.

", "rawMarkdown": "@xhlulu, the link to 256x256 PNG is broken. It says 404 not found.", "votes": 1, "replies": [ { "id": 1316265, "postDate": "2021-05-20T12:18:17.590Z", "content": "

+1 Broken for me as well. I guess its not public yet. Can you please confirm @xhlulu

", "rawMarkdown": "+1 Broken for me as well. I guess its not public yet. Can you please confirm @xhlulu ", "votes": 1 }, { "id": 1316498, "postDate": "2021-05-20T15:45:44.310Z", "content": "

Should work now! Forgot to make it public

", "rawMarkdown": "Should work now! Forgot to make it public", "votes": 2 }, { "id": 1316538, "postDate": "2021-05-20T16:26:56.963Z", "content": "

The last time I used a custom dataset for submission, and it was a week before and private to me. Saved me at the last moment when a teammate replicated my submission and found dataset not found errors.

\n

We made it public and literally saved months of hard work. Phew!

\n

Guess Kaggle should notify when the notebook owner copies the link;

\n
\n

Make sure to make it public before you share the link

\n
", "rawMarkdown": "The last time I used a custom dataset for submission, and it was a week before and private to me. Saved me at the last moment when a teammate replicated my submission and found dataset not found errors.\n\nWe made it public and literally saved months of hard work. Phew!\n\nGuess Kaggle should notify when the notebook owner copies the link;\n> **Make sure to make it public before you share the link**", "votes": 1 }, { "id": 1316688, "postDate": "2021-05-20T19:33:45.003Z", "content": "

Good catch from your teammate; sounds like a almost-nightmare situation :/

\n

A notification would definitely be nice - so easy to miss one when im creating multiple datasets at the same time.

", "rawMarkdown": "Good catch from your teammate; sounds like a almost-nightmare situation :/\n\nA notification would definitely be nice - so easy to miss one when im creating multiple datasets at the same time.", "votes": 2 } ] }, { "id": 1351893, "postDate": "2021-06-16T17:08:14.810Z", "content": "

Thanks so much for this! I am assuming that we will have to manually rescale the bounding boxes to account for the smaller size, right? And any idea what the original dimensions were, so that I can scale by the proper factor?

", "rawMarkdown": "Thanks so much for this! I am assuming that we will have to manually rescale the bounding boxes to account for the smaller size, right? And any idea what the original dimensions were, so that I can scale by the proper factor?", "votes": 2, "replies": [ { "id": 1355848, "postDate": "2021-06-18T15:07:12.513Z", "content": "

Datasets contain a \"meta.csv\" file with the following columns: image_id, dim0, dim1, split.
\ndim0 - seems to be height,
\ndim1 - width.

", "rawMarkdown": "Datasets contain a \"meta.csv\" file with the following columns: image_id, dim0, dim1, split. \ndim0 - seems to be height,\ndim1 - width.", "votes": 2 } ] }, { "id": 1402956, "postDate": "2021-07-28T17:05:50.570Z", "content": "

Hello there, I have a question regarding this. Would someone kindly help me ?
\nSo if we use the resized image for training and testing, when we submit the result, the bounding box coordinates are based on the resized dimension, not the original image dimension. Will this affect our score? Do they compare our results with the ground truth for the original images?

", "rawMarkdown": "Hello there, I have a question regarding this. Would someone kindly help me ? \nSo if we use the resized image for training and testing, when we submit the result, the bounding box coordinates are based on the resized dimension, not the original image dimension. Will this affect our score? Do they compare our results with the ground truth for the original images?", "replies": [ { "id": 1403110, "postDate": "2021-07-28T19:37:53.070Z", "content": "

you can use different image shapes for training and testing..

\n
\n

Do they compare our results with the ground truth for the original images?

\n
\n

yes, the results are evaluated with ground truth labels that are based on the original image shapes - so make sure that you convert correct the bboxes after inference to the original shape.
\nsee public notebooks (with score) to see the details

", "rawMarkdown": "you can use different image shapes for training and testing.. \n>Do they compare our results with the ground truth for the original images?\n\nyes, the results are evaluated with ground truth labels that are based on the original image shapes - so make sure that you convert correct the bboxes after inference to the original shape.\nsee public notebooks (with score) to see the details", "votes": 1 }, { "id": 1403193, "postDate": "2021-07-28T22:32:34.720Z", "content": "

Really clear. Thank you :)

", "rawMarkdown": "Really clear. Thank you :)", "votes": 1 } ] }, { "id": 1378687, "postDate": "2021-07-06T17:50:15.763Z", "content": "

Hi xhlulu,
\nThanks for making the dataset public.
\nHow to can I load the dataset into my notebook , I am trying creating the directory in input folder , but doesn't allow to create.

", "rawMarkdown": "Hi xhlulu,\nThanks for making the dataset public. \nHow to can I load the dataset into my notebook , I am trying creating the directory in input folder , but doesn't allow to create. " }, { "id": 1322690, "postDate": "2021-05-25T15:46:36.560Z", "content": "

How do you know if the VOI LUT information is available or not?
\nSame question for monochrome fixing, is it required for all images or is it a decision that need to be taken case y case?

", "rawMarkdown": "How do you know if the VOI LUT information is available or not?\nSame question for monochrome fixing, is it required for all images or is it a decision that need to be taken case y case?" }, { "id": 1315095, "postDate": "2021-05-19T14:58:26.820Z", "content": "

Just can't believe itπŸ˜‚πŸ˜‚πŸ˜‚Can anyone tell me why the data can be shrink so much----120GB--->nearly 500MB

", "rawMarkdown": "Just can't believe itπŸ˜‚πŸ˜‚πŸ˜‚Can anyone tell me why the data can be shrink so much----120GB--->nearly 500MB", "replies": [ { "id": 1315105, "postDate": "2021-05-19T15:06:22.363Z", "content": "

BTW, why the PNG and JPG can differ so much

", "rawMarkdown": "BTW, why the PNG and JPG can differ so much" }, { "id": 1315207, "postDate": "2021-05-19T16:18:24.167Z", "content": "

If you notice a big difference I recommend using PNG, since JPG compression might not be reliable/great for this sort of use case.

", "rawMarkdown": "If you notice a big difference I recommend using PNG, since JPG compression might not be reliable/great for this sort of use case.", "votes": 2 }, { "id": 1315212, "postDate": "2021-05-19T16:19:10.623Z", "content": "
\n

why the data can be shrink so much

\n
\n

This is becasue the original pixel arrays were very large (e.g. 2000x3000, it really varies)

", "rawMarkdown": "> why the data can be shrink so much\n\nThis is becasue the original pixel arrays were very large (e.g. 2000x3000, it really varies)" } ] }, { "id": 1340102, "postDate": "2021-06-07T17:00:14.617Z", "rawMarkdown": "", "isDeleted": true }, { "id": 1315650, "postDate": "2021-05-20T03:45:20.203Z", "rawMarkdown": "", "isDeleted": true }, { "id": 1324506, "postDate": "2021-05-27T03:05:41.860Z", "content": "

Very helpful, thank you

", "rawMarkdown": "Very helpful, thank you" } ], "comments": [ { "id": 1314809, "author_name": "Shahebaz Mohammad", "author_url": "", "post_date": "2021-05-19T11:49:07.257000", "content": "

The rumour is @xhlulu is working part-time for Kaggle now 😜

\n

This should be great for kagglers like me who are low on bandwidth and GPU all the time ;)

", "votes": 7, "replies": [] }, { "id": 1326630, "author_name": "AnkurSingh", "author_url": "", "post_date": "2021-05-28T15:22:32.160000", "content": "

Hey @xhlulu , thanks for making the dataset public. Just one concern, I went through your preprocessing notebook and it seems that keep_raito=False.

\n

I am getting weird location for bounding boxes. Here, have a look
\n\"Image\"

\n

IMO, the original images are pretty big. So, while reducing the size, aspect ratio & rounding errors will quickly add up.
\nDo let me know what you think? Is anyone else facing the same issue? If no, then its quite possible that I might be doing something wrong πŸ˜…

", "votes": 5, "replies": [ { "id": 1360774, "author_name": "Ayush Thakur", "author_url": "", "post_date": "2021-06-22T10:42:16.630000", "content": "

I haven't faced such an issue so far.

", "votes": 0, "replies": [] } ] }, { "id": 1313116, "author_name": "Ruchi Bhatia", "author_url": "", "post_date": "2021-05-18T12:12:54.590000", "content": "

Hey @xhlulu,
\nReally appreciate you creating the script for converting dicom image data πŸ˜„

\n

I've created a dataset of augmented images based on the jpg files you created!

\n

Augmentation types : RandomSunFlare, RandomFog, RandomBrightness, RandomCrop, Rotate, RGBShift, RandomSnow, HorizontalFlip, VerticalFlip, RandomContrast, HSV

", "votes": 4, "replies": [ { "id": 1313299, "author_name": "Heroseo", "author_url": "", "post_date": "2021-05-18T13:54:10.860000", "content": "

Thanks for sharing. :)
\nAugmented images has also 256 image size?

", "votes": 1, "replies": [] }, { "id": 1313329, "author_name": "Ruchi Bhatia", "author_url": "", "post_date": "2021-05-18T14:10:11.800000", "content": "

Yes @piantic πŸ˜„

", "votes": 1, "replies": [] }, { "id": 1313531, "author_name": "xhlulu", "author_url": "", "post_date": "2021-05-18T15:44:17.670000", "content": "

Cool, thanks for sharing! I think it would be nice if you can share the notebook too so people can try out different configurations of augmentations :)

", "votes": 1, "replies": [] }, { "id": 1313566, "author_name": "Ruchi Bhatia", "author_url": "", "post_date": "2021-05-18T15:57:01.303000", "content": "

I've included that segment in my notebook.
\nHope this helps 😌

", "votes": 3, "replies": [] } ] }, { "id": 1316242, "author_name": "Haswanth Aekula", "author_url": "", "post_date": "2021-05-20T11:51:07.243000", "content": "

@xhlulu, the link to 256x256 PNG is broken. It says 404 not found.

", "votes": 1, "replies": [ { "id": 1316265, "author_name": "Shahebaz Mohammad", "author_url": "", "post_date": "2021-05-20T12:18:17.590000", "content": "

+1 Broken for me as well. I guess its not public yet. Can you please confirm @xhlulu

", "votes": 1, "replies": [] }, { "id": 1316498, "author_name": "xhlulu", "author_url": "", "post_date": "2021-05-20T15:45:44.310000", "content": "

Should work now! Forgot to make it public

", "votes": 2, "replies": [] }, { "id": 1316538, "author_name": "Shahebaz Mohammad", "author_url": "", "post_date": "2021-05-20T16:26:56.963000", "content": "

The last time I used a custom dataset for submission, and it was a week before and private to me. Saved me at the last moment when a teammate replicated my submission and found dataset not found errors.

\n

We made it public and literally saved months of hard work. Phew!

\n

Guess Kaggle should notify when the notebook owner copies the link;

\n
\n

Make sure to make it public before you share the link

\n
", "votes": 1, "replies": [] }, { "id": 1316688, "author_name": "xhlulu", "author_url": "", "post_date": "2021-05-20T19:33:45.003000", "content": "

Good catch from your teammate; sounds like a almost-nightmare situation :/

\n

A notification would definitely be nice - so easy to miss one when im creating multiple datasets at the same time.

", "votes": 2, "replies": [] } ] }, { "id": 1351893, "author_name": "Victor", "author_url": "", "post_date": "2021-06-16T17:08:14.810000", "content": "

Thanks so much for this! I am assuming that we will have to manually rescale the bounding boxes to account for the smaller size, right? And any idea what the original dimensions were, so that I can scale by the proper factor?

", "votes": 2, "replies": [ { "id": 1355848, "author_name": "MikoΕ‚aj Glinka", "author_url": "", "post_date": "2021-06-18T15:07:12.513000", "content": "

Datasets contain a \"meta.csv\" file with the following columns: image_id, dim0, dim1, split.
\ndim0 - seems to be height,
\ndim1 - width.

", "votes": 2, "replies": [] } ] }, { "id": 1402956, "author_name": "Feanor007", "author_url": "", "post_date": "2021-07-28T17:05:50.570000", "content": "

Hello there, I have a question regarding this. Would someone kindly help me ?
\nSo if we use the resized image for training and testing, when we submit the result, the bounding box coordinates are based on the resized dimension, not the original image dimension. Will this affect our score? Do they compare our results with the ground truth for the original images?

", "votes": 0, "replies": [ { "id": 1403110, "author_name": "Ioannis M", "author_url": "", "post_date": "2021-07-28T19:37:53.070000", "content": "

you can use different image shapes for training and testing..

\n
\n

Do they compare our results with the ground truth for the original images?

\n
\n

yes, the results are evaluated with ground truth labels that are based on the original image shapes - so make sure that you convert correct the bboxes after inference to the original shape.
\nsee public notebooks (with score) to see the details

", "votes": 1, "replies": [] }, { "id": 1403193, "author_name": "Feanor007", "author_url": "", "post_date": "2021-07-28T22:32:34.720000", "content": "

Really clear. Thank you :)

", "votes": 1, "replies": [] } ] }, { "id": 1378687, "author_name": "suri@7", "author_url": "", "post_date": "2021-07-06T17:50:15.763000", "content": "

Hi xhlulu,
\nThanks for making the dataset public.
\nHow to can I load the dataset into my notebook , I am trying creating the directory in input folder , but doesn't allow to create.

", "votes": 0, "replies": [] }, { "id": 1322690, "author_name": "Benoit Vallade", "author_url": "", "post_date": "2021-05-25T15:46:36.560000", "content": "

How do you know if the VOI LUT information is available or not?
\nSame question for monochrome fixing, is it required for all images or is it a decision that need to be taken case y case?

", "votes": 0, "replies": [] }, { "id": 1315095, "author_name": "Lucius", "author_url": "", "post_date": "2021-05-19T14:58:26.820000", "content": "

Just can't believe itπŸ˜‚πŸ˜‚πŸ˜‚Can anyone tell me why the data can be shrink so much----120GB--->nearly 500MB

", "votes": 0, "replies": [ { "id": 1315105, "author_name": "Lucius", "author_url": "", "post_date": "2021-05-19T15:06:22.363000", "content": "

BTW, why the PNG and JPG can differ so much

", "votes": 0, "replies": [] }, { "id": 1315207, "author_name": "xhlulu", "author_url": "", "post_date": "2021-05-19T16:18:24.167000", "content": "

If you notice a big difference I recommend using PNG, since JPG compression might not be reliable/great for this sort of use case.

", "votes": 2, "replies": [] }, { "id": 1315212, "author_name": "xhlulu", "author_url": "", "post_date": "2021-05-19T16:19:10.623000", "content": "
\n

why the data can be shrink so much

\n
\n

This is becasue the original pixel arrays were very large (e.g. 2000x3000, it really varies)

", "votes": 0, "replies": [] } ] }, { "id": 1340102, "author_name": "", "author_url": "", "post_date": "2021-06-07T17:00:14.617000", "content": "", "votes": 0, "replies": [] }, { "id": 1315650, "author_name": "", "author_url": "", "post_date": "2021-05-20T03:45:20.203000", "content": "", "votes": 0, "replies": [] }, { "id": 1324506, "author_name": "steven", "author_url": "", "post_date": "2021-05-27T03:05:41.860000", "content": "

Very helpful, thank you

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "1312430": "I resized the images in this competition to JPG and PNG files in different sizes. Before resizing, I apply VOI LUT (see [Raddar's notebook](https://www.kaggle.com/raddar/convert-dicom-to-np-array-the-correct-way)) whenever the information is available (which doesn't seem to be always the case), as well as monochrome fixing (see previous notebook).\n\n[Click here for my preprocessing notebook](https://www.kaggle.com/xhlulu/siim-covid-19-convert-to-jpg)\n\nThe notebook can be easily modified to generate different sizes, aspect ratios and image format. Below is a list of datasets that were created by (variants of) this notebook:\n* [SIIM COVID-19: Resized to 256px JPG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-256px-jpg)\n* [SIIM COVID-19: Resized to 512px JPG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-512px-jpg)\n* [SIIM COVID-19: Resized to 1024px JPG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-1024px-jpg)\n* [SIIM COVID-19: Resized to 224px PNG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-224px-png)\n* [SIIM COVID-19: Resized to 256px PNG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-256px-png)\n* [SIIM COVID-19: Resized to 512px PNG](https://www.kaggle.com/xhlulu/siim-covid19-resized-to-512px-png)", "1314809": "The rumour is @xhlulu is working part-time for Kaggle now 😜\n\nThis should be great for kagglers like me who are low on bandwidth and GPU all the time ;) ", "1326630": "Hey @xhlulu , thanks for making the dataset public. Just one concern, I went through your preprocessing notebook and it seems that `keep_raito=False`. \n\nI am getting weird location for bounding boxes. Here, have a look\n![Image](https://i.imgur.com/TXOkH4y.png)\n\nIMO, the original images are pretty big. So, while reducing the size, aspect ratio & rounding errors will quickly add up. \nDo let me know what you think? Is anyone else facing the same issue? If no, then its quite possible that I might be doing something wrong πŸ˜…\n", "1313116": "Hey @xhlulu,\nReally appreciate you creating the script for converting dicom image data πŸ˜„\n\nI've created a dataset of [augmented images](https://www.kaggle.com/ruchi798/siimfisabiorsna-covid19-detection-augmented) based on the jpg files you created! \n\n**Augmentation types** : RandomSunFlare, RandomFog, RandomBrightness, RandomCrop, Rotate, RGBShift, RandomSnow, HorizontalFlip, VerticalFlip, RandomContrast, HSV", "1316242": "@xhlulu, the link to 256x256 PNG is broken. It says 404 not found.", "1351893": "Thanks so much for this! I am assuming that we will have to manually rescale the bounding boxes to account for the smaller size, right? And any idea what the original dimensions were, so that I can scale by the proper factor?", "1402956": "Hello there, I have a question regarding this. Would someone kindly help me ? \nSo if we use the resized image for training and testing, when we submit the result, the bounding box coordinates are based on the resized dimension, not the original image dimension. Will this affect our score? Do they compare our results with the ground truth for the original images?", "1378687": "Hi xhlulu,\nThanks for making the dataset public. \nHow to can I load the dataset into my notebook , I am trying creating the directory in input folder , but doesn't allow to create. ", "1322690": "How do you know if the VOI LUT information is available or not?\nSame question for monochrome fixing, is it required for all images or is it a decision that need to be taken case y case?", "1315095": "Just can't believe itπŸ˜‚πŸ˜‚πŸ˜‚Can anyone tell me why the data can be shrink so much----120GB--->nearly 500MB", "1340102": "", "1315650": "", "1324506": "Very helpful, thank you" } }