{ "id": 312191, "title": "Score 0.05128 simple mnist model + image processing", "url": "/competitions/ultra-mnist/discussion/312191", "author_name": "Ashish Johnson ", "post_date": "2022-03-10T19:34:35.084000", "votes": 1, "comment_count": 0, "views": 0, "content": "

Main Notebook implementation:> https://www.kaggle.com/ashishrose/score-0-05128-first-ultramnist-model

\n

Steps done:

\n
    \n
  1. Trained a simple MNIST classifier
    2. \n
  2. Took the cleaned dataset from https://www.kaggle.com/remekkinas/ultramnistblack Thanks to Remek Kinas @remekkinas
    \n3.Added contour detection algorithm(Image processing) to extract the numbers from the image.
    3. \n
  3. The detected contours were passed to the trained MNIST classifier to get the output and sum up the output
    4. \n
  4. Logged all the sum in the submission format.
    5. \n
\n

Special Thanks to the below person as the idea of cleaning the data was pretty awesome

\n

LUKASZ BORECKI for this notebook:> https://www.kaggle.com/lukaszborecki/digit-cleaner-concept
\n@lukaszborecki

\n

running end-to-end please refer to last section named Whole inference pipeline end-to-end

\n

Finally, Please Upvote my notebook if you found useful :)

", "messages": [ { "id": 1718400, "postDate": "2022-03-10T19:34:35.083Z", "content": "

Main Notebook implementation:> https://www.kaggle.com/ashishrose/score-0-05128-first-ultramnist-model

\n

Steps done:

\n
    \n
  1. Trained a simple MNIST classifier
    2. \n
  2. Took the cleaned dataset from https://www.kaggle.com/remekkinas/ultramnistblack Thanks to Remek Kinas @remekkinas
    \n3.Added contour detection algorithm(Image processing) to extract the numbers from the image.
    3. \n
  3. The detected contours were passed to the trained MNIST classifier to get the output and sum up the output
    4. \n
  4. Logged all the sum in the submission format.
    5. \n
\n

Special Thanks to the below person as the idea of cleaning the data was pretty awesome

\n

LUKASZ BORECKI for this notebook:> https://www.kaggle.com/lukaszborecki/digit-cleaner-concept
\n@lukaszborecki

\n

running end-to-end please refer to last section named Whole inference pipeline end-to-end

\n

Finally, Please Upvote my notebook if you found useful :)

", "rawMarkdown": "**Main Notebook implementation:>** https://www.kaggle.com/ashishrose/score-0-05128-first-ultramnist-model\n\nSteps done:\n1. Trained a simple MNIST classifier \n2. Took the cleaned dataset from https://www.kaggle.com/remekkinas/ultramnistblack Thanks to [Remek Kinas](https://www.kaggle.com/remekkinas) @remekkinas \n3.Added contour detection algorithm(Image processing) to extract the numbers from the image.\n4. The detected contours were passed to the trained MNIST classifier to get the output and sum up the output\n5. Logged all the sum in the submission format.\n\nSpecial Thanks to the below person as the idea of cleaning the data was pretty awesome\n\n[LUKASZ BORECKI](https://www.kaggle.com/lukaszborecki) for this notebook:> https://www.kaggle.com/lukaszborecki/digit-cleaner-concept\n@lukaszborecki \n\nrunning end-to-end please refer to last section named **Whole inference pipeline end-to-end**\n\n# Finally, Please Upvote my notebook if you found useful :)", "votes": 1 } ], "comments": [], "raw_markdown_by_id": { "1718400": "**Main Notebook implementation:>** https://www.kaggle.com/ashishrose/score-0-05128-first-ultramnist-model\n\nSteps done:\n1. Trained a simple MNIST classifier \n2. Took the cleaned dataset from https://www.kaggle.com/remekkinas/ultramnistblack Thanks to [Remek Kinas](https://www.kaggle.com/remekkinas) @remekkinas \n3.Added contour detection algorithm(Image processing) to extract the numbers from the image.\n4. The detected contours were passed to the trained MNIST classifier to get the output and sum up the output\n5. Logged all the sum in the submission format.\n\nSpecial Thanks to the below person as the idea of cleaning the data was pretty awesome\n\n[LUKASZ BORECKI](https://www.kaggle.com/lukaszborecki) for this notebook:> https://www.kaggle.com/lukaszborecki/digit-cleaner-concept\n@lukaszborecki \n\nrunning end-to-end please refer to last section named **Whole inference pipeline end-to-end**\n\n# Finally, Please Upvote my notebook if you found useful :)" } }