{ "id": 312003, "title": "STEP 1 - Simplify problem - new DATASET - PLAIN BACKGROUND without checkboard", "url": "/competitions/ultra-mnist/discussion/312003", "author_name": "", "post_date": "2022-03-09T23:13:54.017573500Z", "votes": 29, "comment_count": 10, "views": 0, "content": "

First step in my pipeline is …. simplify problem. So I decided to remove background from this game using Digit Cleaner Concept …
\nNew dataset (based on competition data) is ready for download here: ULTRA-MNIST-BLACK

\n

ORYGINAL
\n\"oryg\"

\n

SIMPLIFIED
\n\"Rd\"

\n

\"Re\"

\n

\"Rt\"

\n

To improve:

\n\n

STEP 2 - Find Numbers - NO MODEL REQUIRED … pure CV 💪💪💪

\n

LINK TO DATASET: https://www.kaggle.com/remekkinas/ultramnistblack

", "messages": [ { "id": "1717416", "postDate": "03/09/2022 23:13:54", "content": "

First step in my pipeline is …. simplify problem. So I decided to remove background from this game using Digit Cleaner Concept …
\nNew dataset (based on competition data) is ready for download here: ULTRA-MNIST-BLACK

\n

ORYGINAL
\n\"oryg\"

\n

SIMPLIFIED
\n\"Rd\"

\n

\"Re\"

\n

\"Rt\"

\n

To improve:

\n\n

STEP 2 - Find Numbers - NO MODEL REQUIRED … pure CV 💪💪💪

\n

LINK TO DATASET: https://www.kaggle.com/remekkinas/ultramnistblack

", "rawMarkdown": "First step in my pipeline is .... simplify problem. So I decided to remove background from this game using [Digit Cleaner Concept](https://www.kaggle.com/lukaszborecki/digit-cleaner-concept) ... \n**New dataset (based on competition data) is ready for download here:** [ULTRA-MNIST-BLACK](www.kaggle.com/remekkinas/ultramnistblack)\n\nORYGINAL \n![oryg](https://i.ibb.co/zrgchbt/ultra-mnist-000a.jpg)\n\nSIMPLIFIED\n![Rd](https://i.ibb.co/0DZYLQG/ultra-mnist.jpg)\n\n![Re](https://i.ibb.co/sKxTJGB/ultra-mnist-001.jpg)\n\n![Rt](https://i.ibb.co/GWQD8Yj/ultra-mnist-002.jpg)\n\nTo improve:\n- some (mainly with BIG 0) were not inverted but only few samples (it won't have any impact on score at this moment) - looking for improvement (new release will be available soon)\n\nSTEP 2 - Find Numbers - NO MODEL REQUIRED ... pure CV 💪💪💪\n\n\n**LINK TO DATASET:** https://www.kaggle.com/remekkinas/ultramnistblack", "votes": null }, { "id": "1717516", "postDate": "03/10/2022 01:57:30", "content": "

Hi @remekkinas , thank for the dataset provided.
\nMay I know how do you remove the 'checkboard' background?
\nThanks

", "rawMarkdown": "Hi @remekkinas , thank for the dataset provided. \nMay I know how do you remove the 'checkboard' background?\nThanks", "votes": null }, { "id": "1717518", "postDate": "03/10/2022 02:06:24", "content": "

oh silly me, realized that the simple way is converting the white pixel to black and vice versa?

", "rawMarkdown": "oh silly me, realized that the simple way is converting the white pixel to black and vice versa?", "votes": null }, { "id": "1717620", "postDate": "03/10/2022 04:42:55", "content": "

Thanks so much @remekkinas I used your code and it worked beautifully. I however do not understand how it works. Could you point me to a material to read or a video to watch?
\nThanks

", "rawMarkdown": "Thanks so much @remekkinas I used your code and it worked beautifully. I however do not understand how it works. Could you point me to a material to read or a video to watch? \nThanks", "votes": null }, { "id": "1717782", "postDate": "03/10/2022 08:18:15", "content": "

Overlooked the link you shared, I see you're using Digit cleaner concept notebook created by @lukaszborecki .

", "rawMarkdown": "Overlooked the link you shared, I see you're using Digit cleaner concept notebook created by @lukaszborecki .", "votes": null }, { "id": "1717786", "postDate": "03/10/2022 08:22:47", "content": "

Yes, and created dataset …. to save the planet … no more energy required for doing the same in this competition. I use:

\n\n

and then created public dataset - all of you can use it.

\n

All according to the rules of this competition - dataset can be used.

", "rawMarkdown": "Yes, and created dataset .... to save the planet ... no more energy required for doing the same in this competition. I use:\n- competition dataset\n- notebook which is public and available for use\n\nand then created public dataset - all of you can use it.\n\nAll according to the rules of this competition - dataset can be used.", "votes": null }, { "id": "1718455", "postDate": "03/10/2022 21:00:04", "content": "

I'm new to Computer Vision, hence the naïve question, could you please tell how converting white pixels to black and vice versa helping in removing the 'checkerboard' background?
\nThank you.

", "rawMarkdown": "I'm new to Computer Vision, hence the naïve question, could you please tell how converting white pixels to black and vice versa helping in removing the 'checkerboard' background? \nThank you.", "votes": null }, { "id": "1718460", "postDate": "03/10/2022 21:04:29", "content": "

This is @lukaszborecki notebook. As it appeared we worked on idea in the same time but Lukasz is faster then me … so I decided to use his notebook: https://www.kaggle.com/lukaszborecki/digit-cleaner-concept

", "rawMarkdown": "This is @lukaszborecki notebook. As it appeared we worked on idea in the same time but Lukasz is faster then me ... so I decided to use his notebook: https://www.kaggle.com/lukaszborecki/digit-cleaner-concept", "votes": null }, { "id": "1728582", "postDate": "03/19/2022 03:48:34", "content": "

Hello @jbolaifa , as my understanding, Remek used idea from here.

", "rawMarkdown": "Hello @jbolaifa , as my understanding, Remek used idea from [here](https://www.kaggle.com/code/lukaszborecki/digit-cleaner-concept/notebook).", "votes": null }, { "id": "1728672", "postDate": "03/19/2022 06:35:10", "content": "

Thank you @remekkinas,
\nMay you share source-code that you created this dataset? As my understanding, LUKASZ BORECKI using 2 cleaner method in his code. So how to choose method also threshold value is so useful.

", "rawMarkdown": "Thank you @remekkinas, \nMay you share source-code that you created this dataset? As my understanding, LUKASZ BORECKI using 2 cleaner method in [his code](https://www.kaggle.com/code/lukaszborecki/digit-cleaner-concept/notebook). So how to choose method also threshold value is so useful.", "votes": null }, { "id": "1728790", "postDate": "03/19/2022 09:44:09", "content": "

I used @lukaszborecki script for cleaning out old dataset.
\nNow you have to find different solution for cleaning new dataset. It is still possible.

", "rawMarkdown": "I used @lukaszborecki script for cleaning out old dataset. \nNow you have to find different solution for cleaning new dataset. It is still possible.", "votes": null } ], "comments": [ { "id": 1717516, "author_name": "ivanlaulintiong", "author_url": "", "post_date": "03/10/2022 01:57:30", "content": "

Hi @remekkinas , thank for the dataset provided.
\nMay I know how do you remove the 'checkboard' background?
\nThanks

", "votes": null, "replies": [ { "id": 1717518, "author_name": "ivanlaulintiong", "author_url": "", "post_date": "03/10/2022 02:06:24", "content": "

oh silly me, realized that the simple way is converting the white pixel to black and vice versa?

", "votes": null, "replies": [] }, { "id": 1717782, "author_name": "ivanlaulintiong", "author_url": "", "post_date": "03/10/2022 08:18:15", "content": "

Overlooked the link you shared, I see you're using Digit cleaner concept notebook created by @lukaszborecki .

", "votes": null, "replies": [] }, { "id": 1717786, "author_name": "remekkinas", "author_url": "", "post_date": "03/10/2022 08:22:47", "content": "

Yes, and created dataset …. to save the planet … no more energy required for doing the same in this competition. I use:

\n\n

and then created public dataset - all of you can use it.

\n

All according to the rules of this competition - dataset can be used.

", "votes": null, "replies": [] }, { "id": 1718455, "author_name": "yashikajain", "author_url": "", "post_date": "03/10/2022 21:00:04", "content": "

I'm new to Computer Vision, hence the naïve question, could you please tell how converting white pixels to black and vice versa helping in removing the 'checkerboard' background?
\nThank you.

", "votes": null, "replies": [] }, { "id": 1718460, "author_name": "remekkinas", "author_url": "", "post_date": "03/10/2022 21:04:29", "content": "

This is @lukaszborecki notebook. As it appeared we worked on idea in the same time but Lukasz is faster then me … so I decided to use his notebook: https://www.kaggle.com/lukaszborecki/digit-cleaner-concept

", "votes": null, "replies": [] } ] }, { "id": 1717620, "author_name": "jbolaifa", "author_url": "", "post_date": "03/10/2022 04:42:55", "content": "

Thanks so much @remekkinas I used your code and it worked beautifully. I however do not understand how it works. Could you point me to a material to read or a video to watch?
\nThanks

", "votes": null, "replies": [ { "id": 1728582, "author_name": "phanttan", "author_url": "", "post_date": "03/19/2022 03:48:34", "content": "

Hello @jbolaifa , as my understanding, Remek used idea from here.

", "votes": null, "replies": [] } ] }, { "id": 1728672, "author_name": "phanttan", "author_url": "", "post_date": "03/19/2022 06:35:10", "content": "

Thank you @remekkinas,
\nMay you share source-code that you created this dataset? As my understanding, LUKASZ BORECKI using 2 cleaner method in his code. So how to choose method also threshold value is so useful.

", "votes": null, "replies": [ { "id": 1728790, "author_name": "remekkinas", "author_url": "", "post_date": "03/19/2022 09:44:09", "content": "

I used @lukaszborecki script for cleaning out old dataset.
\nNow you have to find different solution for cleaning new dataset. It is still possible.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1717416": "First step in my pipeline is .... simplify problem. So I decided to remove background from this game using [Digit Cleaner Concept](https://www.kaggle.com/lukaszborecki/digit-cleaner-concept) ... \n**New dataset (based on competition data) is ready for download here:** [ULTRA-MNIST-BLACK](www.kaggle.com/remekkinas/ultramnistblack)\n\nORYGINAL \n![oryg](https://i.ibb.co/zrgchbt/ultra-mnist-000a.jpg)\n\nSIMPLIFIED\n![Rd](https://i.ibb.co/0DZYLQG/ultra-mnist.jpg)\n\n![Re](https://i.ibb.co/sKxTJGB/ultra-mnist-001.jpg)\n\n![Rt](https://i.ibb.co/GWQD8Yj/ultra-mnist-002.jpg)\n\nTo improve:\n- some (mainly with BIG 0) were not inverted but only few samples (it won't have any impact on score at this moment) - looking for improvement (new release will be available soon)\n\nSTEP 2 - Find Numbers - NO MODEL REQUIRED ... pure CV 💪💪💪\n\n\n**LINK TO DATASET:** https://www.kaggle.com/remekkinas/ultramnistblack", "1717516": "Hi @remekkinas , thank for the dataset provided. \nMay I know how do you remove the 'checkboard' background?\nThanks", "1717518": "oh silly me, realized that the simple way is converting the white pixel to black and vice versa?", "1717620": "Thanks so much @remekkinas I used your code and it worked beautifully. I however do not understand how it works. Could you point me to a material to read or a video to watch? \nThanks", "1717782": "Overlooked the link you shared, I see you're using Digit cleaner concept notebook created by @lukaszborecki .", "1717786": "Yes, and created dataset .... to save the planet ... no more energy required for doing the same in this competition. I use:\n- competition dataset\n- notebook which is public and available for use\n\nand then created public dataset - all of you can use it.\n\nAll according to the rules of this competition - dataset can be used.", "1718455": "I'm new to Computer Vision, hence the naïve question, could you please tell how converting white pixels to black and vice versa helping in removing the 'checkerboard' background? \nThank you.", "1718460": "This is @lukaszborecki notebook. As it appeared we worked on idea in the same time but Lukasz is faster then me ... so I decided to use his notebook: https://www.kaggle.com/lukaszborecki/digit-cleaner-concept", "1728582": "Hello @jbolaifa , as my understanding, Remek used idea from [here](https://www.kaggle.com/code/lukaszborecki/digit-cleaner-concept/notebook).", "1728672": "Thank you @remekkinas, \nMay you share source-code that you created this dataset? As my understanding, LUKASZ BORECKI using 2 cleaner method in [his code](https://www.kaggle.com/code/lukaszborecki/digit-cleaner-concept/notebook). So how to choose method also threshold value is so useful.", "1728790": "I used @lukaszborecki script for cleaning out old dataset. \nNow you have to find different solution for cleaning new dataset. It is still possible." }, "source": "meta" }