Link-Cycle GAN to balance imbalanced diseased leaves
\nThis notebook was created by me during participating in the competition Plant Pathology 2020 - FGVC7. I hope this notebook would be helpful for this competition also by some means. Please give it a look and also consider UpVoting if you find the content informative.
\nGiven below are some of the results generated by the cyclicGAN, explained in the above mentioned notebook.
\n![\"image3\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1690611%2Fcf6a02a00265ebc021054119780a8c10%2F__results___45_6.png?generation=1598674000792746&alt=media\")
\n![\"image2\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1690611%2F5e8e460732854c5d614b0b6fc71377ef%2F__results___45_4.png?generation=1598673960255723&alt=media\")
\n![\"imge1\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F1690611%2Fe2d6acc4f9074c04fa27c89274470193%2F__results___44_0.png?generation=1598673879890081&alt=media\")
Link-Cycle GAN to balance imbalanced diseased leaves
\nThis notebook was created by me during participating in the competition Plant Pathology 2020 - FGVC7. I hope this notebook would be helpful for this competition also by some means. Please give it a look and also consider UpVoting if you find the content informative.
\nGiven below are some of the results generated by the cyclicGAN, explained in the above mentioned notebook.
\n
\n
\n
Very nice notebook!!!!
\nInformative.
Thank you very much.
", "rawMarkdown": "Thank you very much.", "votes": null }, { "id": "991665", "postDate": "08/30/2020 15:12:35", "content": "Very good @basu369victor have you compared your implementation of CycleGAN with the provided baseline, to see which perform better?
", "rawMarkdown": "Very good @basu369victor have you compared your implementation of CycleGAN with the provided baseline, to see which perform better?", "votes": null }, { "id": "991797", "postDate": "08/30/2020 17:09:49", "content": "yes, I did. the performance was much lower than the baseline model. but I think If I would have generated a bit high-resolution image with the GAN, I could have outperformed the baseline model.
", "rawMarkdown": "yes, I did. the performance was much lower than the baseline model. but I think If I would have generated a bit high-resolution image with the GAN, I could have outperformed the baseline model.", "votes": null }, { "id": "991821", "postDate": "08/30/2020 17:21:43", "content": "I see @basu369victor , I am doing some research to find ways to improve the CycleGAN baseline, thanks!
", "rawMarkdown": "I see @basu369victor , I am doing some research to find ways to improve the CycleGAN baseline, thanks!", "votes": null }, { "id": "1007357", "postDate": "09/12/2020 05:29:22", "content": "i remember that competition and used SMOTE. However, this is an amazing way to balance the images! Kudos!!
\nWhat impact did it have on the performance?
", "rawMarkdown": "i remember that competition and used SMOTE. However, this is an amazing way to balance the images! Kudos!!\n\nWhat impact did it have on the performance?", "votes": null }, { "id": "1007470", "postDate": "09/12/2020 07:31:01", "content": "It did not go well over the test dataset, I guess generating a bit high-resolution images from gan would have worked
", "rawMarkdown": "It did not go well over the test dataset, I guess generating a bit high-resolution images from gan would have worked", "votes": null } ], "comments": [ { "id": 989787, "author_name": "akshaychavan123", "author_url": "", "post_date": "08/29/2020 05:40:21", "content": "Very nice notebook!!!!
\nInformative.
Thank you very much.
", "votes": null, "replies": [] } ] }, { "id": 991665, "author_name": "dimitreoliveira", "author_url": "", "post_date": "08/30/2020 15:12:35", "content": "Very good @basu369victor have you compared your implementation of CycleGAN with the provided baseline, to see which perform better?
", "votes": null, "replies": [ { "id": 991797, "author_name": "basu369victor", "author_url": "", "post_date": "08/30/2020 17:09:49", "content": "yes, I did. the performance was much lower than the baseline model. but I think If I would have generated a bit high-resolution image with the GAN, I could have outperformed the baseline model.
", "votes": null, "replies": [] }, { "id": 991821, "author_name": "dimitreoliveira", "author_url": "", "post_date": "08/30/2020 17:21:43", "content": "I see @basu369victor , I am doing some research to find ways to improve the CycleGAN baseline, thanks!
", "votes": null, "replies": [] } ] }, { "id": 1007357, "author_name": "fireheart7", "author_url": "", "post_date": "09/12/2020 05:29:22", "content": "i remember that competition and used SMOTE. However, this is an amazing way to balance the images! Kudos!!
\nWhat impact did it have on the performance?
", "votes": null, "replies": [ { "id": 1007470, "author_name": "basu369victor", "author_url": "", "post_date": "09/12/2020 07:31:01", "content": "It did not go well over the test dataset, I guess generating a bit high-resolution images from gan would have worked
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "989703": "Link-[Cycle GAN to balance imbalanced diseased leaves](https://www.kaggle.com/basu369victor/cycle-gan-to-balance-imbalanced-diseased-leaves)\nThis notebook was created by me during participating in the competition [Plant Pathology 2020 - FGVC7](https://www.kaggle.com/c/plant-pathology-2020-fgvc7). I hope this notebook would be helpful for this competition also by some means. Please give it a look and also consider UpVoting if you find the content informative.\nGiven below are some of the results generated by the cyclicGAN, explained in the above mentioned notebook.\n\n\n", "989787": "Very nice notebook!!!!\nInformative.", "989796": "Thank you very much.", "991665": "Very good @basu369victor have you compared your implementation of CycleGAN with the provided baseline, to see which perform better?", "991797": "yes, I did. the performance was much lower than the baseline model. but I think If I would have generated a bit high-resolution image with the GAN, I could have outperformed the baseline model.", "991821": "I see @basu369victor , I am doing some research to find ways to improve the CycleGAN baseline, thanks!", "1007357": "i remember that competition and used SMOTE. However, this is an amazing way to balance the images! Kudos!!\n\nWhat impact did it have on the performance?", "1007470": "It did not go well over the test dataset, I guess generating a bit high-resolution images from gan would have worked" }, "source": "meta" }