π Algovera: Algorithms for Detecting and Anonymizing Faces
This is a collaboration with DataUnion to develop a face anonymisation algorithm.
Who is DataUnion?
DataUnion creates data unions to crowdsource datasets and give ownership to the data creators, annotators and verifiers. Unlike platforms like Amazon Mechanical Turk where contributors are paid a small amount per image, DataUnion instead gives contributors the option to co-own the machine-learning-ready dataset. The contributors can exchange these shares (datatokens) for a stable currency, they can stake the datatokens to get continuous rewards when the dataset is being sold, or they can hold on to the tokens in the expectation that the value of the dataset will increase with future R&D. The created dataset is listed and made available for purchase on a marketplace such as Ocean. Here, investors who think the dataset has value and utility can purchase the datatoken, stake into the dataset, or earn interest by exchanging data tokens with other currencies.
Why is this important?
Distributing ownership of a dataset among creators prevents one centralised party from controlling it. In this framework, the value of any algorithms built on top of the dataset can flow back to the community rather than a single entity. Furthermore, the algorithms themselves (or maybe even the apps built on top) can be owned by the community. We are getting close to an ImageNet moment for AI in Web3. For the first time, we can implement a Proof of Concept for decentralized AI - one that is created and controlled by the community rather than Big Tech.
Contents
πͺ Community
Hacking Sessions
- Algovera hacking sessions: Fridays at 11:30 UTC. 30 min. Anyone is welcome to drop in!
π - Catch up on previous sessions in the Hacking Sessions playlist of our YouTube
- Join the discussion in the
π -face-anonymisation channel of our Discord
Algovera
Algovera is a community of individuals working to facilitate and accelerate the development of decentralised AI products and research.
Website | Notion | Discord | Calendar | Twitter | YouTube
π Initial Setup
Set up environment
Open a new terminal and:
#clone repo
git clone https://github.com/AlgoveraAI/face-anonymization.git
cd face-anonymizationπ€ Resources
- Introductory Presentation
- Resources Notebook
π License
The license is MIT. Details