Hi, I've also made a utility script and a usage demonstration kernel with some useful functions that I'm using on this competition, There are some generic functions, some related to image segmentation, and some specific to this competition and Keras.
\n\nA few things that liked about this new feature, how you can iterate faster and have a cleaner and more robust code, also I think that two kinds of scrips will be especially useful, one kind with functions related to competition and another kind with more generic functions.
\n\nScript kernel: cloud images segmentation utillity script\nUsage kernel: Cloud Segmentation with utility scripts and Keras
\n\nWhat you will find on the script I made:\n- All used dependencies\n- External repository codes (need internet option ON)\n- Seed function (to make model runs more reproducible)\n- Segmentation functions related to this competition\n- Multi-thread data process functions (to resize and apply transformations faster)\n- Model evaluation (training plots)\n- Model post-process (Set threshold and removing small masks)\n- Prediction evaluation (Generate metrics over predictions and sample evaluation)\n- Data generator\n- Learning rate schedulers
\n\nIf anyone else also made public utility scripts for this competition please let us know.
", "messages": [ { "id": "642969", "postDate": "10/06/2019 21:49:41", "content": "Hi, I've also made a utility script and a usage demonstration kernel with some useful functions that I'm using on this competition, There are some generic functions, some related to image segmentation, and some specific to this competition and Keras.
\n\nA few things that liked about this new feature, how you can iterate faster and have a cleaner and more robust code, also I think that two kinds of scrips will be especially useful, one kind with functions related to competition and another kind with more generic functions.
\n\nScript kernel: cloud images segmentation utillity script\nUsage kernel: Cloud Segmentation with utility scripts and Keras
\n\nWhat you will find on the script I made:\n- All used dependencies\n- External repository codes (need internet option ON)\n- Seed function (to make model runs more reproducible)\n- Segmentation functions related to this competition\n- Multi-thread data process functions (to resize and apply transformations faster)\n- Model evaluation (training plots)\n- Model post-process (Set threshold and removing small masks)\n- Prediction evaluation (Generate metrics over predictions and sample evaluation)\n- Data generator\n- Learning rate schedulers
\n\nIf anyone else also made public utility scripts for this competition please let us know.
", "rawMarkdown": "Hi, I've also made a [utility script](https://www.kaggle.com/dimitreoliveira/cloud-images-segmentation-utillity-script) and a [usage demonstration](https://www.kaggle.com/dimitreoliveira/cloud-segmentation-with-utility-scripts-and-keras) kernel with some useful functions that I'm using on this competition, There are some generic functions, some related to image segmentation, and some specific to this competition and Keras.\n\nA few things that liked about this new feature, how you can iterate faster and have a cleaner and more robust code, also I think that two kinds of scrips will be especially useful, one kind with functions related to competition and another kind with more generic functions.\n\nScript kernel: [cloud images segmentation utillity script](https://www.kaggle.com/dimitreoliveira/cloud-images-segmentation-utillity-script)\nUsage kernel: [Cloud Segmentation with utility scripts and Keras](https://www.kaggle.com/dimitreoliveira/cloud-segmentation-with-utility-scripts-and-keras)\n\n**What you will find on the script I made:**\n- All used dependencies\n- External repository codes (need internet option ON)\n- Seed function (to make model runs more reproducible)\n- Segmentation functions related to this competition\n- Multi-thread data process functions (to resize and apply transformations faster)\n- Model evaluation (training plots)\n- Model post-process (Set threshold and removing small masks)\n- Prediction evaluation (Generate metrics over predictions and sample evaluation)\n- Data generator\n- Learning rate schedulers\n\nIf anyone else also made public utility scripts for this competition please let us know.", "votes": null }, { "id": "643005", "postDate": "10/06/2019 23:27:30", "content": "Cool! Very useful.
", "rawMarkdown": "Cool! Very useful.", "votes": null }, { "id": "643021", "postDate": "10/07/2019 00:49:56", "content": "Thak you
", "rawMarkdown": "Thak you", "votes": null }, { "id": "643096", "postDate": "10/07/2019 04:55:25", "content": "Very Helpful.... Thanks @dimitreoliveira
", "rawMarkdown": "Very Helpful.... Thanks @dimitreoliveira", "votes": null }, { "id": "643122", "postDate": "10/07/2019 06:21:20", "content": "Great write-up\nThanks for sharing @dimitreoliveira
", "rawMarkdown": "Great write-up\nThanks for sharing @dimitreoliveira", "votes": null } ], "comments": [ { "id": 643005, "author_name": "gogo827jz", "author_url": "", "post_date": "10/06/2019 23:27:30", "content": "Cool! Very useful.
", "votes": null, "replies": [] }, { "id": 643021, "author_name": "jmourad100", "author_url": "", "post_date": "10/07/2019 00:49:56", "content": "Thak you
", "votes": null, "replies": [] }, { "id": 643096, "author_name": "veeralakrishna", "author_url": "", "post_date": "10/07/2019 04:55:25", "content": "Very Helpful.... Thanks @dimitreoliveira
", "votes": null, "replies": [] }, { "id": 643122, "author_name": "chirag9073", "author_url": "", "post_date": "10/07/2019 06:21:20", "content": "Great write-up\nThanks for sharing @dimitreoliveira
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "642969": "Hi, I've also made a [utility script](https://www.kaggle.com/dimitreoliveira/cloud-images-segmentation-utillity-script) and a [usage demonstration](https://www.kaggle.com/dimitreoliveira/cloud-segmentation-with-utility-scripts-and-keras) kernel with some useful functions that I'm using on this competition, There are some generic functions, some related to image segmentation, and some specific to this competition and Keras.\n\nA few things that liked about this new feature, how you can iterate faster and have a cleaner and more robust code, also I think that two kinds of scrips will be especially useful, one kind with functions related to competition and another kind with more generic functions.\n\nScript kernel: [cloud images segmentation utillity script](https://www.kaggle.com/dimitreoliveira/cloud-images-segmentation-utillity-script)\nUsage kernel: [Cloud Segmentation with utility scripts and Keras](https://www.kaggle.com/dimitreoliveira/cloud-segmentation-with-utility-scripts-and-keras)\n\n**What you will find on the script I made:**\n- All used dependencies\n- External repository codes (need internet option ON)\n- Seed function (to make model runs more reproducible)\n- Segmentation functions related to this competition\n- Multi-thread data process functions (to resize and apply transformations faster)\n- Model evaluation (training plots)\n- Model post-process (Set threshold and removing small masks)\n- Prediction evaluation (Generate metrics over predictions and sample evaluation)\n- Data generator\n- Learning rate schedulers\n\nIf anyone else also made public utility scripts for this competition please let us know.", "643005": "Cool! Very useful.", "643021": "Thak you", "643096": "Very Helpful.... Thanks @dimitreoliveira", "643122": "Great write-up\nThanks for sharing @dimitreoliveira" }, "source": "meta" }