url
stringlengths 52
64
| url_highres
stringlengths 54
66
| url_source
stringlengths 50
57
|
---|---|---|
https://www.flickr.com/photos/37161495@N00/24197167113 |
||
https://www.flickr.com/photos/145599221@N04/38226150665 |
||
https://www.flickr.com/photos/44324716@N08/46081239324 |
||
https://www.flickr.com/photos/137520364@N06/42320287484 |
||
https://www.flickr.com/photos/145660676@N03/33526239656 |
||
https://www.flickr.com/photos/145615154@N06/30144620671 |
||
https://www.flickr.com/photos/94771018@N03/51108025087 |
||
https://www.flickr.com/photos/148748355@N05/48051936773 |
||
https://www.flickr.com/photos/165415395@N07/49384394963 |
||
https://www.flickr.com/photos/40231929@N02/18617603329 |
||
https://www.flickr.com/photos/147330136@N06/32973385864 |
||
https://www.flickr.com/photos/76858203@N04/21070900323 |
||
https://www.flickr.com/photos/84947370@N00/24394783021 |
||
https://www.flickr.com/photos/59081381@N03/46436662045 |
||
https://www.flickr.com/photos/129713999@N05/26246845344 |
||
https://www.flickr.com/photos/63999290@N00/47961445567 |
||
https://www.flickr.com/photos/110141246@N07/20683452365 |
||
https://www.flickr.com/photos/56007743@N08/51466861018 |
||
https://www.flickr.com/photos/130953354@N08/32856612131 |
||
https://www.flickr.com/photos/108605878@N06/15041414305 |
||
https://www.flickr.com/photos/184155407@N04/48714103966 |
||
https://www.flickr.com/photos/122076518@N08/26738456364 |
||
https://www.flickr.com/photos/68842444@N03/9626583020 |
||
https://www.flickr.com/photos/70719205@N05/7880478010 |
||
https://www.flickr.com/photos/76111143@N02/40058100220 |
||
https://www.flickr.com/photos/150111929@N04/39051707874 |
||
https://www.flickr.com/photos/112254964@N05/50795028522 |
||
https://www.flickr.com/photos/29468641@N04/16860332149 |
||
https://www.flickr.com/photos/26246711@N00/2147962834 |
||
https://www.flickr.com/photos/90589513@N00/16794848601 |
||
https://www.flickr.com/photos/59076512@N05/50299189031 |
||
https://www.flickr.com/photos/133616431@N06/48778047468 |
||
https://www.flickr.com/photos/135957213@N07/50674308328 |
||
https://www.flickr.com/photos/11314862@N02/35755971615 |
||
https://www.flickr.com/photos/190749011@N04/50529715717 |
||
https://www.flickr.com/photos/164632874@N04/51252362027 |
||
https://www.flickr.com/photos/44840123@N07/18622052335 |
||
https://www.flickr.com/photos/150111929@N04/51483842919 |
||
https://www.flickr.com/photos/92947007@N03/51578730647 |
||
https://www.flickr.com/photos/155509408@N02/44129786422 |
||
https://www.flickr.com/photos/8101022@N05/6272618537 |
||
https://www.flickr.com/photos/127819143@N02/51674246253 |
||
https://www.flickr.com/photos/25397255@N05/43050263420 |
||
https://www.flickr.com/photos/166279667@N06/48094240062 |
||
https://www.flickr.com/photos/155509408@N02/45256448485 |
||
https://www.flickr.com/photos/41555360@N03/32011310437 |
||
https://www.flickr.com/photos/101561334@N08/22799345819 |
||
https://www.flickr.com/photos/94966166@N02/26581793784 |
||
https://www.flickr.com/photos/191419476@N04/50935046358 |
||
https://www.flickr.com/photos/60054626@N00/23748778694 |
||
https://www.flickr.com/photos/34302538@N06/5033906375 |
||
https://www.flickr.com/photos/96064256@N04/45815200654 |
||
https://www.flickr.com/photos/91501748@N07/43647879715 |
||
https://www.flickr.com/photos/59927874@N06/32923894653 |
||
https://www.flickr.com/photos/101005855@N06/28424811357 |
||
https://www.flickr.com/photos/127906254@N06/29858033212 |
||
https://www.flickr.com/photos/29470713@N04/34402938462 |
||
https://www.flickr.com/photos/121262492@N02/45655637271 |
||
https://www.flickr.com/photos/61609021@N07/48138322378 |
||
https://www.flickr.com/photos/48125931@N00/9880851136 |
||
https://www.flickr.com/photos/49487266@N07/4558101291 |
||
https://www.flickr.com/photos/149756499@N02/44209460684 |
||
https://www.flickr.com/photos/158730994@N05/45506897725 |
||
https://www.flickr.com/photos/110227198@N03/50726736848 |
||
https://www.flickr.com/photos/95929490@N06/21725422743 |
||
https://www.flickr.com/photos/38345529@N05/12489550714 |
||
https://www.flickr.com/photos/58585543@N02/50881067928 |
||
https://www.flickr.com/photos/128788011@N04/29286237997 |
||
https://www.flickr.com/photos/133697664@N06/51668963738 |
||
https://www.flickr.com/photos/94771018@N03/28889415732 |
||
https://www.flickr.com/photos/97592948@N07/40886018525 |
||
https://www.flickr.com/photos/131902054@N02/20029172974 |
||
https://www.flickr.com/photos/70125105@N06/35068791193 |
||
https://www.flickr.com/photos/135565977@N04/24004873386 |
||
https://www.flickr.com/photos/141459056@N08/41729502375 |
||
https://www.flickr.com/photos/76111143@N02/48168998781 |
||
https://www.flickr.com/photos/55188902@N08/37680166922 |
||
https://www.flickr.com/photos/62309569@N03/45575225271 |
||
https://www.flickr.com/photos/53714574@N06/7488055194 |
||
https://www.flickr.com/photos/126115381@N08/21415824890 |
||
https://www.flickr.com/photos/11807887@N07/4179507705 |
||
https://www.flickr.com/photos/149271704@N02/50389132872 |
||
https://www.flickr.com/photos/101561334@N08/16058567421 |
||
https://www.flickr.com/photos/64165273@N07/34724878720 |
||
https://www.flickr.com/photos/65509508@N03/41153742270 |
||
https://www.flickr.com/photos/133414628@N03/49869801507 |
||
https://www.flickr.com/photos/134168349@N08/25825255751 |
||
https://www.flickr.com/photos/47430793@N08/45656294195 |
||
https://www.flickr.com/photos/138224835@N02/43927005802 |
||
https://www.flickr.com/photos/152183014@N04/32797198732 |
||
https://www.flickr.com/photos/148592346@N07/49032950717 |
||
https://www.flickr.com/photos/97195226@N03/26001721565 |
||
https://www.flickr.com/photos/34586311@N05/6360162863 |
||
https://www.flickr.com/photos/20696196@N08/16755987448 |
||
https://www.flickr.com/photos/191419476@N04/50787228903 |
||
https://www.flickr.com/photos/45717276@N08/49229784701 |
||
https://www.flickr.com/photos/59081381@N03/33654636408 |
||
https://www.flickr.com/photos/121396426@N06/37733747341 |
||
https://www.flickr.com/photos/147773026@N05/46809658111 |
||
https://www.flickr.com/photos/46907600@N02/6867228457 |
🗿 Megalith-10m
What is Megalith-10m?
Megalith-10m is a dataset of ~10 million links to Flickr images that were categorized as "photo" with license info of:
- No known copyright restrictions (Flickr commons), or
- United States Government Work, or
- Public Domain Dedication (CC0), or
- Public Domain Mark
What's the intended use of Megalith-10m?
Megalith-10m is intended to contain only links to wholesome unedited uncopyrighted photographs - the sort of images that we humans see when we walk around outside. I collected Megalith-10m for the purpose of training neural networks, but you're welcome to use Megalith-10m for whatever you want. Of course, I recommend conducting your own independent analysis of content and copyright status before using Megalith-linked images in Serious Projects.
Where can I get text captions for Megalith-10m?
- DrawThings.ai have uploaded
megalith-10m-sharecap
(captions made with ShareCaptioner) - AI Picasso have uploaded
megalith-10m-florence2
(captions made with Florence 2) - CaptionEmporium have uploaded
flickr-megalith-10m-internvl2-multi-caption
(captions made with InternVL2-8B as well as shorter single-sentence captions made by summarizing the InternVL2/Florence2/ShareCaptioner results with Llama3.1-8B) - DrawThings.ai is working on further captioning with MoonDream2
How can I efficiently download the images referenced by Megalith-10m?
- DrawThings.ai has archived the images linked by Megalith-10m here: https://huggingface.co/datasets/drawthingsai/megalith-10m
- If you want to download Megalith-10m images directly from Flickr, I posted a sample downloading command you can use with img2dataset
How was Megalith-10m collected?
I used the Flickr API to query for photos matching some basic criteria (SFW photo with CC0 / public domain license info), which gave me around 12 million links. I then used various filtering strategies to exclude ~2m image links which didn't appear to point to wholesome public-domain minimally-edited photos. These filtering strategies included:
- Account-level filtering, based on
- Manual adjudication for the top 5000 most prolific accounts
- Repeated-watermark detection
- Photo-level filtering, based on
Image metadata
- Mention of copyright restrictions in the EXIF tags
- Mention of copyright restrictions in the text description
Image content
Duplicate detection
CLIP-assisted checking for
- Clearly non-photo images (illustrations, screenshots, 3d renders, etc.)
- Clearly non-wholesome images (violence, nudity, etc.)
Minimum-resolution enforcement (at least 256x256 pixels)
Manual spot-checking of some images and metadata
What content does Megalith-10m contain?
The demo notebook shows a random sample of 100 images being loaded from the links in Megalith-10m.
Based on this random sample, I would estimate the following dataset statistics:
- 5-7% of images may have minor edits or annotations (timestamps, color grading, borders, etc.)
- 1-2% of images may be copyright-constrained (watermarks or text descriptions cast doubt on the license metadata)
- 1-2% of images may be non-wholesome (guns, suggestive poses, etc.)
- 1-2% of images may be non-photos (paintings, screenshots, etc.)
Is 10 million images really enough to teach a neural network about the visual world?
For the parts of the visual world that are well-represented in Megalith-10m, definitely! Projects like CommonCanvas, Mitsua Diffusion, and Matryoshka Diffusion have shown that you can train useable generative models on similarly-sized image datasets. Of course, many parts of the world aren't well-represented in Megalith-10m, so you'd need additional data to learn about those.
What have people done with Megalith-10m?
- AI Picasso have successfully trained a full text-to-image model CommonArt β on Megalith-10m (and other open datasets).
- I've successfully trained small text-to-image models on Megalith-10m for my own education.
- Megalith-10m was among the datasets used to train Janus, DeepSeek's AR model for multimodal understanding and generation
- Downloads last month
- 520
Data Sourcing report
No elements in this dataset have been identified as either opted-out, or opted-in, by their creator.