I'm currently working on uploading more than 70k high resolution images that were retrieved from Flickr by Nvidia for their StyleGAN paper. I believe this might be a good complement to the 1-million fake faces dataset uploaded by Bojan.
\n\nHere they are:\n* Part 1\n* Part 2\n* Part 3\n* Part 4\n* Part 5\n* Part 6\n* Part 7\n* Part 8\n* Part 9
\n\nYou can find a version of all 70k images resized to 256x256, and compressed in JPEG, which was done using this notebook.
\n\nHere is the description on the official Github:
\n\n> Flickr-Faces-HQ (FFHQ) is a high-quality image dataset of human faces, originally created as a benchmark for generative adversarial networks (GAN):
\n\n> A Style-Based Generator Architecture for Generative Adversarial Networks\n> Tero Karras (NVIDIA), Samuli Laine (NVIDIA), Timo Aila (NVIDIA)\n> https://arxiv.org/abs/1812.04948
\n\n> The dataset consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos.
", "messages": [ { "id": 700963, "postDate": "2019-12-22T22:02:17.010Z", "content": "I'm currently working on uploading more than 70k high resolution images that were retrieved from Flickr by Nvidia for their StyleGAN paper. I believe this might be a good complement to the 1-million fake faces dataset uploaded by Bojan.
\n\nHere they are:\n* Part 1\n* Part 2\n* Part 3\n* Part 4\n* Part 5\n* Part 6\n* Part 7\n* Part 8\n* Part 9
\n\nYou can find a version of all 70k images resized to 256x256, and compressed in JPEG, which was done using this notebook.
\n\nHere is the description on the official Github:
\n\n> Flickr-Faces-HQ (FFHQ) is a high-quality image dataset of human faces, originally created as a benchmark for generative adversarial networks (GAN):
\n\n> A Style-Based Generator Architecture for Generative Adversarial Networks\n> Tero Karras (NVIDIA), Samuli Laine (NVIDIA), Timo Aila (NVIDIA)\n> https://arxiv.org/abs/1812.04948
\n\n> The dataset consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos.
", "rawMarkdown": "I'm currently working on uploading more than 70k high resolution images that were [retrieved from Flickr by Nvidia](https://github.com/NVlabs/ffhq-dataset) for their StyleGAN paper. I believe this might be a good complement to the 1-million fake faces dataset uploaded by Bojan.\n\nHere they are:\n* [Part 1](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-1)\n* [Part 2](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-2)\n* [Part 3](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-3)\n* [Part 4](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-4)\n* [Part 5](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-5)\n* [Part 6](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-6)\n* [Part 7](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-7)\n* [Part 8](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-8)\n* [Part 9](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-9)\n\nYou can find a [version of all 70k images](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px) resized to 256x256, and compressed in JPEG, which was done using [this notebook](https://www.kaggle.com/xhlulu/resize-flickr-70k-to-256x256/output).\n\n## Context\n\nHere is the description on the official Github:\n\n> Flickr-Faces-HQ (FFHQ) is a high-quality image dataset of human faces, originally created as a benchmark for generative adversarial networks (GAN):\n\n> A Style-Based Generator Architecture for Generative Adversarial Networks\n> Tero Karras (NVIDIA), Samuli Laine (NVIDIA), Timo Aila (NVIDIA)\n> https://arxiv.org/abs/1812.04948\n\n> The dataset consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos.", "votes": 53 }, { "id": 743623, "postDate": "2020-02-12T06:30:05.440Z", "content": "@xhlulu Thanks for the dataset! This is really precious especially concerning that extracted frames have much more fake than real face images. Really appreciate how you made 256x dataset public! Saves lot of time & computation
\n\nI find the page 404, though. Any help appreciated\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px
", "rawMarkdown": "@xhlulu Thanks for the dataset! This is really precious especially concerning that extracted frames have much more fake than real face images. Really appreciate how you made 256x dataset public! Saves lot of time & computation\n\nI find the page 404, though. Any help appreciated\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px", "votes": 1, "replies": [ { "id": 748196, "postDate": "2020-02-17T09:10:49.263Z", "content": "I have the same problem as you :( Did you manage to download it?
", "rawMarkdown": "I have the same problem as you :( Did you manage to download it?\n" }, { "id": 748197, "postDate": "2020-02-17T09:12:34.140Z", "content": "@roguekk007 Just managed to download it. Instead of using the link to the dataset, you should just open the kernel used to create the resized version, and download the resized data directly from there. Hope this helps.
", "rawMarkdown": "@roguekk007 Just managed to download it. Instead of using the link to the dataset, you should just open the kernel used to create the resized version, and download the resized data directly from there. Hope this helps.", "votes": 1 }, { "id": 748212, "postDate": "2020-02-17T09:33:45.453Z", "content": "@ngcferreira Been away from this competition for a while. Thanks for the tip!
", "rawMarkdown": "@ngcferreira Been away from this competition for a while. Thanks for the tip!" }, { "id": 748456, "postDate": "2020-02-17T14:36:47.357Z", "content": "Sorry I forgot to make it public, it should work now
", "rawMarkdown": "Sorry I forgot to make it public, it should work now" }, { "id": 748523, "postDate": "2020-02-17T16:13:16.050Z", "content": "@xhlulu Thanks for the dataset!!
", "rawMarkdown": "@xhlulu Thanks for the dataset!!", "votes": 1 } ] }, { "id": 703306, "postDate": "2019-12-26T01:36:59.790Z", "content": "Here's 70k more from that dataset, all 128x128 thumbnail images: https://www.kaggle.com/greatgamedota/ffhq-face-data-set
", "rawMarkdown": "Here's 70k more from that dataset, all 128x128 thumbnail images: https://www.kaggle.com/greatgamedota/ffhq-face-data-set", "votes": 2 }, { "id": 702907, "postDate": "2019-12-25T10:34:14.273Z", "content": "Thanks for your uploading, this competition's data are really large, I need better computer.😂
", "rawMarkdown": "Thanks for your uploading, this competition's data are really large, I need better computer.😂 ", "votes": 2, "replies": [ { "id": 703164, "postDate": "2019-12-25T18:05:04.630Z", "content": "If you downsize the images you should be able to fit it in a kaggle kernel ;)
", "rawMarkdown": "If you downsize the images you should be able to fit it in a kaggle kernel ;)" } ] }, { "id": 2309363, "postDate": "2023-06-19T15:05:52.857Z", "content": "Im curious to understand why you grouped the pictures into so many folders and subfolders instead of one folder with all the picture? I cant figure out what practical reason you did that, please share some insight. It is harder to now extract all the images into one file.
", "rawMarkdown": "Im curious to understand why you grouped the pictures into so many folders and subfolders instead of one folder with all the picture? I cant figure out what practical reason you did that, please share some insight. It is harder to now extract all the images into one file. " }, { "id": 1623726, "postDate": "2021-12-20T07:12:03.850Z", "content": "thanks for sharing
\nare the images of the dataset free of copyright permission for publishing research in journals
well done
", "rawMarkdown": "well done" }, { "id": 743895, "postDate": "2020-02-12T11:25:33.783Z", "content": "@xhlulu would u have updated dataset link for these images\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px ?
", "rawMarkdown": "@xhlulu would u have updated dataset link for these images\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px ?\n\n" }, { "id": 727369, "postDate": "2020-01-23T17:02:38.477Z", "content": "Nice thanks.. @xhlulu \nare these extracted from Kaggle set ?\n2) So this will help model in learning more about the features of an original face compared to distorted one ?
", "rawMarkdown": "Nice thanks.. @xhlulu \nare these extracted from Kaggle set ?\n2) So this will help model in learning more about the features of an original face compared to distorted one ?" }, { "id": 777995, "postDate": "2020-03-18T04:15:04.503Z", "rawMarkdown": "", "isDeleted": true, "replies": [ { "id": 778008, "postDate": "2020-03-18T04:31:20.917Z", "content": "It would be against the rules of the competition because some of the video sources have licenses that is restricted to non-commercial use only.
", "rawMarkdown": "It would be against the rules of the competition because some of the video sources have licenses that is restricted to non-commercial use only.", "votes": 1 }, { "id": 778023, "postDate": "2020-03-18T04:51:51.417Z", "rawMarkdown": "", "isDeleted": true }, { "id": 781340, "postDate": "2020-03-21T07:04:36.707Z", "rawMarkdown": "", "isDeleted": true } ] }, { "id": 703720, "postDate": "2019-12-26T15:04:27.533Z", "content": "Thanks!
", "rawMarkdown": "Thanks!", "votes": 1 }, { "id": 703539, "postDate": "2019-12-26T09:26:31.123Z", "content": "Thanks for sharing!
", "rawMarkdown": "Thanks for sharing!", "votes": 1 }, { "id": 705849, "postDate": "2019-12-29T15:17:33.683Z", "content": "Nice thanks!
", "rawMarkdown": "Nice thanks!", "votes": 2 }, { "id": 701102, "postDate": "2019-12-23T04:52:26.447Z", "content": "Great\nThanks for Sharing @xhlulu
", "rawMarkdown": "Great\nThanks for Sharing @xhlulu " }, { "id": 917219, "postDate": "2020-07-06T10:27:22.503Z", "content": "thanks for sharing!!!
", "rawMarkdown": "thanks for sharing!!!" }, { "id": 727026, "postDate": "2020-01-23T11:47:32.027Z", "content": "Thanks for sharing!
", "rawMarkdown": "Thanks for sharing!" } ], "comments": [ { "id": 743623, "author_name": "Nicholas Lyu", "author_url": "", "post_date": "2020-02-12T06:30:05.440000", "content": "@xhlulu Thanks for the dataset! This is really precious especially concerning that extracted frames have much more fake than real face images. Really appreciate how you made 256x dataset public! Saves lot of time & computation
\n\nI find the page 404, though. Any help appreciated\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px
", "votes": 1, "replies": [ { "id": 748196, "author_name": "Nuno Ferreira", "author_url": "", "post_date": "2020-02-17T09:10:49.263000", "content": "I have the same problem as you :( Did you manage to download it?
", "votes": 0, "replies": [] }, { "id": 748197, "author_name": "Nuno Ferreira", "author_url": "", "post_date": "2020-02-17T09:12:34.140000", "content": "@roguekk007 Just managed to download it. Instead of using the link to the dataset, you should just open the kernel used to create the resized version, and download the resized data directly from there. Hope this helps.
", "votes": 1, "replies": [] }, { "id": 748212, "author_name": "Nicholas Lyu", "author_url": "", "post_date": "2020-02-17T09:33:45.453000", "content": "@ngcferreira Been away from this competition for a while. Thanks for the tip!
", "votes": 0, "replies": [] }, { "id": 748456, "author_name": "xhlulu", "author_url": "", "post_date": "2020-02-17T14:36:47.357000", "content": "Sorry I forgot to make it public, it should work now
", "votes": 0, "replies": [] }, { "id": 748523, "author_name": "Debanga Raj Neog", "author_url": "", "post_date": "2020-02-17T16:13:16.050000", "content": "@xhlulu Thanks for the dataset!!
", "votes": 1, "replies": [] } ] }, { "id": 703306, "author_name": "GreatGameDota", "author_url": "", "post_date": "2019-12-26T01:36:59.790000", "content": "Here's 70k more from that dataset, all 128x128 thumbnail images: https://www.kaggle.com/greatgamedota/ffhq-face-data-set
", "votes": 2, "replies": [] }, { "id": 702907, "author_name": "DiegoJohnson", "author_url": "", "post_date": "2019-12-25T10:34:14.273000", "content": "Thanks for your uploading, this competition's data are really large, I need better computer.😂
", "votes": 2, "replies": [ { "id": 703164, "author_name": "xhlulu", "author_url": "", "post_date": "2019-12-25T18:05:04.630000", "content": "If you downsize the images you should be able to fit it in a kaggle kernel ;)
", "votes": 0, "replies": [] } ] }, { "id": 2309363, "author_name": "ZakirC", "author_url": "", "post_date": "2023-06-19T15:05:52.857000", "content": "Im curious to understand why you grouped the pictures into so many folders and subfolders instead of one folder with all the picture? I cant figure out what practical reason you did that, please share some insight. It is harder to now extract all the images into one file.
", "votes": 0, "replies": [] }, { "id": 1623726, "author_name": "ihsan sahib", "author_url": "", "post_date": "2021-12-20T07:12:03.850000", "content": "thanks for sharing
\nare the images of the dataset free of copyright permission for publishing research in journals
well done
", "votes": 0, "replies": [] }, { "id": 743895, "author_name": "Jaideep", "author_url": "", "post_date": "2020-02-12T11:25:33.783000", "content": "@xhlulu would u have updated dataset link for these images\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px ?
", "votes": 0, "replies": [] }, { "id": 727369, "author_name": "Jaideep", "author_url": "", "post_date": "2020-01-23T17:02:38.477000", "content": "Nice thanks.. @xhlulu \nare these extracted from Kaggle set ?\n2) So this will help model in learning more about the features of an original face compared to distorted one ?
", "votes": 0, "replies": [] }, { "id": 777995, "author_name": "", "author_url": "", "post_date": "2020-03-18T04:15:04.503000", "content": "", "votes": 0, "replies": [ { "id": 778008, "author_name": "David", "author_url": "", "post_date": "2020-03-18T04:31:20.917000", "content": "It would be against the rules of the competition because some of the video sources have licenses that is restricted to non-commercial use only.
", "votes": 1, "replies": [] }, { "id": 778023, "author_name": "", "author_url": "", "post_date": "2020-03-18T04:51:51.417000", "content": "", "votes": 0, "replies": [] }, { "id": 781340, "author_name": "", "author_url": "", "post_date": "2020-03-21T07:04:36.707000", "content": "", "votes": 0, "replies": [] } ] }, { "id": 703720, "author_name": "Agastya Kommanamanchi", "author_url": "", "post_date": "2019-12-26T15:04:27.533000", "content": "Thanks!
", "votes": 1, "replies": [] }, { "id": 703539, "author_name": "plot", "author_url": "", "post_date": "2019-12-26T09:26:31.123000", "content": "Thanks for sharing!
", "votes": 1, "replies": [] }, { "id": 705849, "author_name": "Carlo", "author_url": "", "post_date": "2019-12-29T15:17:33.683000", "content": "Nice thanks!
", "votes": 2, "replies": [] }, { "id": 701102, "author_name": "Ailurophile", "author_url": "", "post_date": "2019-12-23T04:52:26.447000", "content": "Great\nThanks for Sharing @xhlulu
", "votes": 0, "replies": [] }, { "id": 917219, "author_name": "Jackling_Gu", "author_url": "", "post_date": "2020-07-06T10:27:22.503000", "content": "thanks for sharing!!!
", "votes": 0, "replies": [] }, { "id": 727026, "author_name": "Hieu Phung", "author_url": "", "post_date": "2020-01-23T11:47:32.027000", "content": "Thanks for sharing!
", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "700963": "I'm currently working on uploading more than 70k high resolution images that were [retrieved from Flickr by Nvidia](https://github.com/NVlabs/ffhq-dataset) for their StyleGAN paper. I believe this might be a good complement to the 1-million fake faces dataset uploaded by Bojan.\n\nHere they are:\n* [Part 1](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-1)\n* [Part 2](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-2)\n* [Part 3](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-3)\n* [Part 4](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-4)\n* [Part 5](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-5)\n* [Part 6](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-6)\n* [Part 7](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-7)\n* [Part 8](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-8)\n* [Part 9](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-part-9)\n\nYou can find a [version of all 70k images](https://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px) resized to 256x256, and compressed in JPEG, which was done using [this notebook](https://www.kaggle.com/xhlulu/resize-flickr-70k-to-256x256/output).\n\n## Context\n\nHere is the description on the official Github:\n\n> Flickr-Faces-HQ (FFHQ) is a high-quality image dataset of human faces, originally created as a benchmark for generative adversarial networks (GAN):\n\n> A Style-Based Generator Architecture for Generative Adversarial Networks\n> Tero Karras (NVIDIA), Samuli Laine (NVIDIA), Timo Aila (NVIDIA)\n> https://arxiv.org/abs/1812.04948\n\n> The dataset consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos.", "743623": "@xhlulu Thanks for the dataset! This is really precious especially concerning that extracted frames have much more fake than real face images. Really appreciate how you made 256x dataset public! Saves lot of time & computation\n\nI find the page 404, though. Any help appreciated\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px", "703306": "Here's 70k more from that dataset, all 128x128 thumbnail images: https://www.kaggle.com/greatgamedota/ffhq-face-data-set", "702907": "Thanks for your uploading, this competition's data are really large, I need better computer.😂 ", "2309363": "Im curious to understand why you grouped the pictures into so many folders and subfolders instead of one folder with all the picture? I cant figure out what practical reason you did that, please share some insight. It is harder to now extract all the images into one file. ", "1623726": "thanks for sharing\nare the images of the dataset free of copyright permission for publishing research in journals", "752679": "well done", "743895": "@xhlulu would u have updated dataset link for these images\nhttps://www.kaggle.com/xhlulu/flickrfaceshq-dataset-nvidia-resized-256px ?\n\n", "727369": "Nice thanks.. @xhlulu \nare these extracted from Kaggle set ?\n2) So this will help model in learning more about the features of an original face compared to distorted one ?", "777995": "", "703720": "Thanks!", "703539": "Thanks for sharing!", "705849": "Nice thanks!", "701102": "Great\nThanks for Sharing @xhlulu ", "917219": "thanks for sharing!!!", "727026": "Thanks for sharing!" } }