{ "id": 304993, "title": "GAN Training Procedure - Make unlimited COTS", "url": "/competitions/tensorflow-great-barrier-reef/discussion/304993", "author_name": "", "post_date": "2022-02-03T10:01:04.855090700Z", "votes": 8, "comment_count": 3, "views": 0, "content": "

Dear Kagglers,

\n

As promised I have prepared the GAN training procedure and also the BBox ETL where you are able to train you own GAN from scratch. I have tried to explain many parts of code to allow you understand what is going on :)
\nGan Training notebook
\nGAN Generating notebook
\nIf you have any suggestions feel free :)

", "messages": [ { "id": "1674207", "postDate": "02/03/2022 10:01:04", "content": "

Dear Kagglers,

\n

As promised I have prepared the GAN training procedure and also the BBox ETL where you are able to train you own GAN from scratch. I have tried to explain many parts of code to allow you understand what is going on :)
\nGan Training notebook
\nGAN Generating notebook
\nIf you have any suggestions feel free :)

", "rawMarkdown": "Dear Kagglers,\n\nAs promised I have prepared the GAN training procedure and also the BBox ETL where you are able to train you own GAN from scratch. I have tried to explain many parts of code to allow you understand what is going on :) \n[Gan Training notebook](https://www.kaggle.com/marcinstasko/gan-training-make-unlimited-cots)\n[GAN Generating notebook](https://www.kaggle.com/marcinstasko/unlimited-cots-generator-pytorch-gan-in-action)\nIf you have any suggestions feel free :)", "votes": null }, { "id": "1674393", "postDate": "02/03/2022 12:58:04", "content": "

Pardon me if my question is stupid but can I ask how to use these data in order to train yolo models?

", "rawMarkdown": "Pardon me if my question is stupid but can I ask how to use these data in order to train yolo models?", "votes": null }, { "id": "1674405", "postDate": "02/03/2022 13:12:41", "content": "

@locbaop there are no stupid questions :)
\nHow you can apply GAN augmentation technique:
\n1) Train the Generator (explained here)
\n2) Generate new samples (explained here, you can skip first part because you can use model I shared)
\n3) Make data augmentation (paste generated cots onto picture without cots)

\n

When people will be interested in these work I will consider to prepare the notebook where I will use generated snippets to augument the data :)

", "rawMarkdown": "locbaop there are no stupid questions :)\nHow you can apply GAN augmentation technique:\n1) Train the Generator ([explained here](https://www.kaggle.com/marcinstasko/gan-training-make-unlimited-cots)) \n2) Generate new samples ([explained here](https://www.kaggle.com/marcinstasko/unlimited-cots-generator-pytorch-gan-in-action), you can skip first part because you can use model I [shared](https://www.kaggle.com/marcinstasko/cots-ganv1))\n3) Make data augmentation (paste generated cots onto picture without cots)\n\nWhen people will be interested in these work I will consider to prepare the notebook where I will use generated snippets to augument the data :)", "votes": null }, { "id": "1674421", "postDate": "02/03/2022 13:35:24", "content": "

thank you @marcinstasko , we're waiting for your augment data notebook. ❤️

", "rawMarkdown": "thank you @marcinstasko , we're waiting for your augment data notebook. ❤️", "votes": null } ], "comments": [ { "id": 1674393, "author_name": "locbaop", "author_url": "", "post_date": "02/03/2022 12:58:04", "content": "

Pardon me if my question is stupid but can I ask how to use these data in order to train yolo models?

", "votes": null, "replies": [ { "id": 1674405, "author_name": "marcinstasko", "author_url": "", "post_date": "02/03/2022 13:12:41", "content": "

@locbaop there are no stupid questions :)
\nHow you can apply GAN augmentation technique:
\n1) Train the Generator (explained here)
\n2) Generate new samples (explained here, you can skip first part because you can use model I shared)
\n3) Make data augmentation (paste generated cots onto picture without cots)

\n

When people will be interested in these work I will consider to prepare the notebook where I will use generated snippets to augument the data :)

", "votes": null, "replies": [] }, { "id": 1674421, "author_name": "locbaop", "author_url": "", "post_date": "02/03/2022 13:35:24", "content": "

thank you @marcinstasko , we're waiting for your augment data notebook. ❤️

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1674207": "Dear Kagglers,\n\nAs promised I have prepared the GAN training procedure and also the BBox ETL where you are able to train you own GAN from scratch. I have tried to explain many parts of code to allow you understand what is going on :) \n[Gan Training notebook](https://www.kaggle.com/marcinstasko/gan-training-make-unlimited-cots)\n[GAN Generating notebook](https://www.kaggle.com/marcinstasko/unlimited-cots-generator-pytorch-gan-in-action)\nIf you have any suggestions feel free :)", "1674393": "Pardon me if my question is stupid but can I ask how to use these data in order to train yolo models?", "1674405": "locbaop there are no stupid questions :)\nHow you can apply GAN augmentation technique:\n1) Train the Generator ([explained here](https://www.kaggle.com/marcinstasko/gan-training-make-unlimited-cots)) \n2) Generate new samples ([explained here](https://www.kaggle.com/marcinstasko/unlimited-cots-generator-pytorch-gan-in-action), you can skip first part because you can use model I [shared](https://www.kaggle.com/marcinstasko/cots-ganv1))\n3) Make data augmentation (paste generated cots onto picture without cots)\n\nWhen people will be interested in these work I will consider to prepare the notebook where I will use generated snippets to augument the data :)", "1674421": "thank you @marcinstasko , we're waiting for your augment data notebook. ❤️" }, "source": "meta" }