I'm excited to be announcing our newest TPU Stars, awarded to the most knowledgeable and helpful TPU experts in this competition. This particular competition was more challenging, so a big congrats to these winners who were able to demonstrate the ability to tackle this tough task of creating generative images using TPUs!
\n✨ DimitreOliveira wrote 2 complementary notebooks. First an Introduction to CycleGAN - Monet Paintings which provided a great baseline for getting started on this problem. And then the follow-on Improving CycleGAN - Monet Paintings full of experimentations to advance generative approaches. He created and shared these TFRecords from an additional set of Berkeley images. Dimitre has also been very helpful in his forum interactions, providing insights and replies to fellow community members. @dimitreoliveira
\n✨ Jesper Sören Dramsch introduced us to Generative Adversarial Networks on TPUs with his Getting started with standard GANs - Tutorial. Then supplemented that intro with Understanding and Improving CycleGANs - Tutorial, providing additional approaches to creating GANs. @jesperdramsch
Thanks, Dimitre & Jesper! Both of these stars will be seeing their 20 hrs/wk of additional TPU quota for each of the next four weeks. Please join me in congratulating them!
", "messages": [ { "id": "1089756", "postDate": "11/24/2020 18:40:54", "content": "I'm excited to be announcing our newest TPU Stars, awarded to the most knowledgeable and helpful TPU experts in this competition. This particular competition was more challenging, so a big congrats to these winners who were able to demonstrate the ability to tackle this tough task of creating generative images using TPUs!
\n✨ DimitreOliveira wrote 2 complementary notebooks. First an Introduction to CycleGAN - Monet Paintings which provided a great baseline for getting started on this problem. And then the follow-on Improving CycleGAN - Monet Paintings full of experimentations to advance generative approaches. He created and shared these TFRecords from an additional set of Berkeley images. Dimitre has also been very helpful in his forum interactions, providing insights and replies to fellow community members. @dimitreoliveira
\n✨ Jesper Sören Dramsch introduced us to Generative Adversarial Networks on TPUs with his Getting started with standard GANs - Tutorial. Then supplemented that intro with Understanding and Improving CycleGANs - Tutorial, providing additional approaches to creating GANs. @jesperdramsch
Thanks, Dimitre & Jesper! Both of these stars will be seeing their 20 hrs/wk of additional TPU quota for each of the next four weeks. Please join me in congratulating them!
", "rawMarkdown": "I'm excited to be announcing our newest [TPU Stars](https://www.kaggle.com/tpu-stars), awarded to the most knowledgeable and helpful TPU experts in this competition. This particular competition was more challenging, so a big congrats to these winners who were able to demonstrate the ability to tackle this tough task of creating generative images using TPUs!\n\n✨ [**DimitreOliveira**](https://www.kaggle.com/dimitreoliveira) wrote 2 complementary notebooks. First an [Introduction to CycleGAN - Monet Paintings](https://www.kaggle.com/dimitreoliveira/introduction-to-cyclegan-monet-paintings) which provided a great baseline for getting started on this problem. And then the follow-on [Improving CycleGAN - Monet Paintings](https://www.kaggle.com/dimitreoliveira/improving-cyclegan-monet-paintings) full of experimentations to advance generative approaches. He created and shared [these TFRecords from an additional set of Berkeley images](https://www.kaggle.com/dimitreoliveira/monet-paintings-berkeley-tfrecords-256x256). Dimitre has also been very helpful in his forum interactions, providing insights and replies to fellow community members. @dimitreoliveira \n✨ [**Jesper Sören Dramsch**](https://www.kaggle.com/jesperdramsch) introduced us to Generative Adversarial Networks on TPUs with his [Getting started with standard GANs - Tutorial](https://www.kaggle.com/jesperdramsch/getting-started-with-standard-gans-tutorial). Then supplemented that intro with [Understanding and Improving CycleGANs - Tutorial](https://www.kaggle.com/jesperdramsch/understanding-and-improving-cyclegans-tutorial), providing additional approaches to creating GANs. @jesperdramsch \n\nThanks, Dimitre & Jesper! Both of these stars will be seeing their 20 hrs/wk of additional TPU quota for each of the next four weeks. Please join me in congratulating them!", "votes": null }, { "id": "1089848", "postDate": "11/24/2020 20:45:39", "content": "Thanks, @juliaelliott , and congratulation @jesperdramsch !
\nFor quite some time I wanted to take some time and do experiments with GANs, this competition was a good opportunity.
\nA funny thing is that one of the first times I thought that \"deep learning was close to black magic\", was the time that I saw CycleGAN tuning a horse to zebra on a video, was cool to do some experimentations with it 😄.
Wow! Thanks, @juliaelliott! Congrats @dimitreoliveira!
\nThese are two of my less popular notebooks, so this is a fantastic surprise! Looking forward to playing around with the new TPU Pods.
Congratulations! @dimitreoliveira and @jesperdramsch and thanks @juliaelliott for encouraging such informative contributions
", "rawMarkdown": "Congratulations! @dimitreoliveira and @jesperdramsch and thanks @juliaelliott for encouraging such informative contributions", "votes": null }, { "id": "3237032", "postDate": "06/30/2025 20:34:36", "content": "Congratulations 😜, it's been 5 years🙃
", "rawMarkdown": "Congratulations 😜, it's been 5 years🙃", "votes": null } ], "comments": [ { "id": 1089848, "author_name": "dimitreoliveira", "author_url": "", "post_date": "11/24/2020 20:45:39", "content": "Thanks, @juliaelliott , and congratulation @jesperdramsch !
\nFor quite some time I wanted to take some time and do experiments with GANs, this competition was a good opportunity.
\nA funny thing is that one of the first times I thought that \"deep learning was close to black magic\", was the time that I saw CycleGAN tuning a horse to zebra on a video, was cool to do some experimentations with it 😄.
Wow! Thanks, @juliaelliott! Congrats @dimitreoliveira!
\nThese are two of my less popular notebooks, so this is a fantastic surprise! Looking forward to playing around with the new TPU Pods.
Congratulations! @dimitreoliveira and @jesperdramsch and thanks @juliaelliott for encouraging such informative contributions
", "votes": null, "replies": [] }, { "id": 3237032, "author_name": "sanjidh090", "author_url": "", "post_date": "06/30/2025 20:34:36", "content": "Congratulations 😜, it's been 5 years🙃
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1089756": "I'm excited to be announcing our newest [TPU Stars](https://www.kaggle.com/tpu-stars), awarded to the most knowledgeable and helpful TPU experts in this competition. This particular competition was more challenging, so a big congrats to these winners who were able to demonstrate the ability to tackle this tough task of creating generative images using TPUs!\n\n✨ [**DimitreOliveira**](https://www.kaggle.com/dimitreoliveira) wrote 2 complementary notebooks. First an [Introduction to CycleGAN - Monet Paintings](https://www.kaggle.com/dimitreoliveira/introduction-to-cyclegan-monet-paintings) which provided a great baseline for getting started on this problem. And then the follow-on [Improving CycleGAN - Monet Paintings](https://www.kaggle.com/dimitreoliveira/improving-cyclegan-monet-paintings) full of experimentations to advance generative approaches. He created and shared [these TFRecords from an additional set of Berkeley images](https://www.kaggle.com/dimitreoliveira/monet-paintings-berkeley-tfrecords-256x256). Dimitre has also been very helpful in his forum interactions, providing insights and replies to fellow community members. @dimitreoliveira \n✨ [**Jesper Sören Dramsch**](https://www.kaggle.com/jesperdramsch) introduced us to Generative Adversarial Networks on TPUs with his [Getting started with standard GANs - Tutorial](https://www.kaggle.com/jesperdramsch/getting-started-with-standard-gans-tutorial). Then supplemented that intro with [Understanding and Improving CycleGANs - Tutorial](https://www.kaggle.com/jesperdramsch/understanding-and-improving-cyclegans-tutorial), providing additional approaches to creating GANs. @jesperdramsch \n\nThanks, Dimitre & Jesper! Both of these stars will be seeing their 20 hrs/wk of additional TPU quota for each of the next four weeks. Please join me in congratulating them!", "1089848": "Thanks, @juliaelliott , and congratulation @jesperdramsch !\nFor quite some time I wanted to take some time and do experiments with GANs, this competition was a good opportunity.\nA funny thing is that one of the first times I thought that \"deep learning was close to black magic\", was the time that I saw CycleGAN tuning a horse to zebra on a video, was cool to do some experimentations with it 😄.", "1091238": "Wow! Thanks, @juliaelliott! Congrats @dimitreoliveira!\nThese are two of my less popular notebooks, so this is a fantastic surprise! Looking forward to playing around with the new TPU Pods.", "2170091": "Congratulations! @dimitreoliveira and @jesperdramsch and thanks @juliaelliott for encouraging such informative contributions", "3237032": "Congratulations 😜, it's been 5 years🙃" }, "source": "meta" }