Computer Vision for All (Developers) ?
What if this new self-serve platform was not an internal tool, but a platform opened for all developers?
There are already some computer vision web-services out there, but pictures, moving or still, are core to every experience of the Facebook products galaxy. With billions of them transiting through these products every day, Facebook has the raw material and the business motivation to develop the most sophisticated picture analysis artificial intelligence possible.
Those AIs contribute a lot to a secure (forbidden contents identification), optimized (photos framing, subjects identification) and accessible (rich description of what is in a photo for visually impaired people) user experience on Facebook.
Facebook already provides such open access to its AI algorithms through Wit.ai, for natural language processing. By opening access to its computer vision IA to all developers, Facebook could improve its computer vision capabilities. Each application would be an opportunity to train its IA to recognize new objects categories, emotions, actions…
Let’s imagine that we will be able to use this AI this April. At what use could we put it? Here are some quick ideas:
- A nutrition assistant bot: I send a picture of everything I eat to the bot, and it produces reports about the quality of my food habits. When in doubt, it asks, and I contribute to the IA training by indicating the kind of food in the picture, and the quantity of it.
- A “Photos Missions” app: Make sure your kid spend more productive time with your iPhone by looking around for the picture they have to take. The app sends them on missions (take a picture of something blue, an animal, a book) and automatically confirms their success. When in doubt, a parent contributes to the IA training by establishing if the mission is accomplished or not.
What would be your ideas?