This is very interesting, literally the day I do a talk on CycleGANs and introduce a package for it, a competition for similar image translation problems is launched!
\nPlease check out my talk on @abhishek's channel over here where I discuss how CycleGANs work, the various pitfalls and failures of CycleGANs, and the implementation with my library UPIT.
\nAdditionally, I have written a starter kernel that demonstrates how easy it is to train a CycleGAN model.
\nI didn't get a chance to work on this earlier and when I did, I made a dumb mistake while copying and pasting code LOL so just fixed it all and it should work now :)
\nPlease let me know if you have any suggestions. Other model implementations will be added soon.
", "messages": [ { "id": "993503", "postDate": "09/01/2020 01:29:30", "content": "This is very interesting, literally the day I do a talk on CycleGANs and introduce a package for it, a competition for similar image translation problems is launched!
\nPlease check out my talk on @abhishek's channel over here where I discuss how CycleGANs work, the various pitfalls and failures of CycleGANs, and the implementation with my library UPIT.
\nAdditionally, I have written a starter kernel that demonstrates how easy it is to train a CycleGAN model.
\nI didn't get a chance to work on this earlier and when I did, I made a dumb mistake while copying and pasting code LOL so just fixed it all and it should work now :)
\nPlease let me know if you have any suggestions. Other model implementations will be added soon.
", "rawMarkdown": "This is very interesting, literally the day I do a talk on CycleGANs and introduce a package for it, a competition for similar image translation problems is launched!\n\nPlease check out my talk on @abhishek's channel over [here](https://www.youtube.com/watch?v=gT8-wDPLOBg) where I discuss how CycleGANs work, the various pitfalls and failures of CycleGANs, and the implementation with my library [UPIT](https://github.com/tmabraham/UPIT). \n\nAdditionally, I have written a starter [kernel](https://www.kaggle.com/tanlikesmath/upit-a-package-for-unpaired-img2img-translation) that demonstrates how easy it is to train a CycleGAN model. \n\nI didn't get a chance to work on this earlier and when I did, I made a dumb mistake while copying and pasting code LOL so just fixed it all and it should work now :)\n\nPlease let me know if you have any suggestions. Other model implementations will be added soon.", "votes": null }, { "id": "994775", "postDate": "09/02/2020 00:01:33", "content": "Hey, @tanlikesmath that was a very great talk, The CycleGAN introduction was brief, but it was a very good presentation, You are doing very interesting research, good luck. Also, you have quite a good CV, congratulations!
", "rawMarkdown": "Hey, @tanlikesmath that was a very great talk, The CycleGAN introduction was brief, but it was a very good presentation, You are doing very interesting research, good luck. Also, you have quite a good CV, congratulations!", "votes": null }, { "id": "994782", "postDate": "09/02/2020 00:10:58", "content": "Thank you for your feedback!
\nYes, the CycleGAN introduction was shorter than I would have liked, but there was a lot to cover! I might write a longer blog post in the future. Nevertheless, I am glad you enjoyed it!
", "rawMarkdown": "Thank you for your feedback!\n\nYes, the CycleGAN introduction was shorter than I would have liked, but there was a lot to cover! I might write a longer blog post in the future. Nevertheless, I am glad you enjoyed it!", "votes": null }, { "id": "994793", "postDate": "09/02/2020 00:36:58", "content": "Yeah indeed that is a long topic, that blog post would be nice, especially if covers your application on medicine, going through your modification from the vanilla CycleGAN, what worked and what did not, if you write it, make sure to post a link at this competition.
", "rawMarkdown": "Yeah indeed that is a long topic, that blog post would be nice, especially if covers your application on medicine, going through your modification from the vanilla CycleGAN, what worked and what did not, if you write it, make sure to post a link at this competition.", "votes": null }, { "id": "994928", "postDate": "09/02/2020 04:31:09", "content": "Thank you for the suggestions! I will make sure to post it over here! :)
", "rawMarkdown": "Thank you for the suggestions! I will make sure to post it over here! :)", "votes": null } ], "comments": [ { "id": 994775, "author_name": "dimitreoliveira", "author_url": "", "post_date": "09/02/2020 00:01:33", "content": "Hey, @tanlikesmath that was a very great talk, The CycleGAN introduction was brief, but it was a very good presentation, You are doing very interesting research, good luck. Also, you have quite a good CV, congratulations!
", "votes": null, "replies": [ { "id": 994782, "author_name": "tanlikesmath", "author_url": "", "post_date": "09/02/2020 00:10:58", "content": "Thank you for your feedback!
\nYes, the CycleGAN introduction was shorter than I would have liked, but there was a lot to cover! I might write a longer blog post in the future. Nevertheless, I am glad you enjoyed it!
", "votes": null, "replies": [] }, { "id": 994793, "author_name": "dimitreoliveira", "author_url": "", "post_date": "09/02/2020 00:36:58", "content": "Yeah indeed that is a long topic, that blog post would be nice, especially if covers your application on medicine, going through your modification from the vanilla CycleGAN, what worked and what did not, if you write it, make sure to post a link at this competition.
", "votes": null, "replies": [] }, { "id": 994928, "author_name": "tanlikesmath", "author_url": "", "post_date": "09/02/2020 04:31:09", "content": "Thank you for the suggestions! I will make sure to post it over here! :)
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "993503": "This is very interesting, literally the day I do a talk on CycleGANs and introduce a package for it, a competition for similar image translation problems is launched!\n\nPlease check out my talk on @abhishek's channel over [here](https://www.youtube.com/watch?v=gT8-wDPLOBg) where I discuss how CycleGANs work, the various pitfalls and failures of CycleGANs, and the implementation with my library [UPIT](https://github.com/tmabraham/UPIT). \n\nAdditionally, I have written a starter [kernel](https://www.kaggle.com/tanlikesmath/upit-a-package-for-unpaired-img2img-translation) that demonstrates how easy it is to train a CycleGAN model. \n\nI didn't get a chance to work on this earlier and when I did, I made a dumb mistake while copying and pasting code LOL so just fixed it all and it should work now :)\n\nPlease let me know if you have any suggestions. Other model implementations will be added soon.", "994775": "Hey, @tanlikesmath that was a very great talk, The CycleGAN introduction was brief, but it was a very good presentation, You are doing very interesting research, good luck. Also, you have quite a good CV, congratulations!", "994782": "Thank you for your feedback!\n\nYes, the CycleGAN introduction was shorter than I would have liked, but there was a lot to cover! I might write a longer blog post in the future. Nevertheless, I am glad you enjoyed it!", "994793": "Yeah indeed that is a long topic, that blog post would be nice, especially if covers your application on medicine, going through your modification from the vanilla CycleGAN, what worked and what did not, if you write it, make sure to post a link at this competition.", "994928": "Thank you for the suggestions! I will make sure to post it over here! :)" }, "source": "meta" }