AutoSight

Earn income while you drive. Autosight is a metagraph-powered app that allows users to capture images of the road, and sell them directly to self-driving car companies for AI model training.

Comment 10

A Need for Data

Artificial intelligence companies are in constant need for more and more data to help train their models as the AI race heats up. Models are growing in size and capabilities, but can become limited in effectiveness without a large amount of data to learn from.

Specifically, self-driving AI companies building autonomous vehicles rely on image datasets to train their vision models to detect objects in the real world. These models must detect vehicles, pedestrians, signs, and any other objects captured from a vehicle's cameras.

To ensure the best performance, the imagery dataset needs to be:

  • Large: Tens of thousands of images are needed to adequately provide examples of all the different scenarios seen on the road.
  • Diverse: Images must be captured at a variety of times, in a variety of weather conditions, at different roadways and locations around the world.
  • Viewpoint-Specific: Images must be captured from a vehicle's point of view, so objects are viewed from the same perspective as the vehicle's cameras.

AI companies are actively looking to purchase imagery datasets that meet these requirements, but they do not exist.

A Data Market Solution

To address this unmet need for data, AutoSight was developed.

AutoSight is a system that enables users to earn income while they drive, by contributing to a dataset of road images that AI companies pay to access.

Users use the AutoSight mobile app to work as a dashboard camera that automatically captures and uploads images of the road, earning them AUTO tokens. AI companies purchase access to these images from the AutoSight web app, paying for access with AUTO tokens.

This data marketplace benefits both the users, or 'sellers', and AI companies, or 'buyers', of this image data.

Users gain a new avenue to become a "datapreneur" that passively earns extra income by collecting and providing high-demand data. AI companies gain access to an ever-growing dataset of quality images they can use to train their models.

How It Works

AutoSight is comprised of a mobile app, a web app, and a metagraph backend.

Mobile App: Users download the AutoSight app to their phone, and can use it every time they drive.

With the app open, users input their wallet address, then follow instructions to set up their phone to work as a dashboard camera, positioning it so the camera is centered on the road ahead of the vehicle.

From here, users can start the capture process and begin driving. The app will automatically collect images of the road, along with the current time and location. This information is uploaded to the metagraph, and users receive 10 AUTO tokens for each upload.

Users simply drive as normal, passively earning rewards along the way. When they reach their destination, they can stop the capture and view their total AUTO rewards.

Web App: AI companies seeking access to images can browse to the AutoSight web app, called the AutoSight Explorer.

The app interface shows an interactive map that plots the locations of all the road images uploaded by users and available for purchase. Previews of each image are also shown, allowing customers to view their content and quality.

With further development, the app will include wallet integration to enable the purchasing of image access. Customers will use AUTO tokens to access image download options, allowing them to bulk donwload images for training.

Search options will be added to filter images by capture time and location, so AI companies can target images colleted at specific times and geographic regions.

Metagraph: The metagraph works as the blockchain backend that ties the mobile app and web app together.

It is responsible for validating, storing, and publishing road image data records.

The metagraph accepts new image data sent from users' AutoSight app. It checks this data to validate its capture dates and times, and stores it within its On Chain state. Ten AUTO tokens are then minted on the metagraph for the user's reward address.

The metagraph includes custom endpoint routes that provide access to its state. The AutoSight Explorer makes requests to these endpoints to collect and display the image data within its interface for customers to view.

In the future, the metagraph will accept new transactions that include an AUTO payment, to update image access controls for customers' accounts.

Tokenomics

AUTO is the native token for the AutoSight metagraph.

These tokens are used to reward users for contributing road images to the metagraph's dataset, and will be required by customers to purchase access to these images.

Users' wallet addresses are minted 10 AUTO tokens for each image they upload to the metagraph. On the other hand, it will cost 10 AUTO tokens to purchase access to download a single image. AUTO tokens used to purchase image access will be burned and taken out of circulation.

The cycle of minting AUTO rewards, then burning them when used will help to stabilize the circulation and prevent inflation.

Tokens can be freely traded and swapped on the open market, using marketplaces like UniSwap.

The demand for AUTO tokens from AI companies will be met by the supply of tokens from users. This demand will drive the USD value of AUTO tokens, allowing users to turn their rewards into dollars.

Build Process

The AutoSight metagraph was developed using the Euclid Development Environment.

The Mobile App was built using Android Studio, and the Web app was built using React.

Built With

+ 36 more
Share this project:

Updates