Megalith-10m
Megalith-10m is a dataset composed of approximately 10 million links to unmodified Flickr images from free sources such as the public domain or the American government. Perfect for training diverse visual models.
Description
The dataset Megalith-10m contains nearly 10 million links to Flickr photographs that are labeled royalty-free, including public domain and US government works. These images represent varied scenes of daily life and nature.
What is this dataset for?
- Train computer vision models on a variety of natural images
- Test the robustness of models on real and diverse scenes
- Establishing a basis for image recognition research
Can it be enriched or improved?
Yes, it is possible to add additional annotations or to download images to create specific subsets. A verification of content and metadata is recommended for professional use.
🔎 In summary
🧠 Recommended for
- Vision researchers
- AI developers
- Natural image analysis projects
🔧 Compatible tools
- PyTorch
- TensorFlow
- Hugging Face datasets
- Download scripts
💡 Tip
Prepare a download and filtering pipeline to effectively use this dataset.
Frequently Asked Questions
Does the dataset contain the images directly?
No, it only contains links to images hosted on Flickr.
Can I use the images for commercial use?
Yes, the images are under free licenses compatible with commercial use (MIT).
Do you have to check the rights of the images individually?
It is recommended to do an independent verification of rights for sensitive or commercial projects.
