I had the pleasure meeting Mohammad Musa, CEO deepen AI, within Silicon Valley. Deepen is an AI to:
- Calibrate
- Annotate
- Validate
Their speciallity is automotive, but I do belive this AI is convertable for further scenarios. Geographically the’re located in Asia, US and Germany (Europe).
Calibration
For calibration Deepen AI uses Safety Pool, a platform shared by stakeholders, such as :
- OEMs & ADAS Developers
- Governments & Regulators
- Tier 1s & L2 + tech. providers
- Researchers
This covers all relevant parties for end-to-end delivering of services; also for highly regulated markets, such as automotive.
Sensor calibration itself is time-consuming and regularly needed, e.g. when switching front-shields of modern cars (360° view camera), auto-light, auto-whipers, advanced cruising controls etc. Regular checks and adjustments take place due to rough roads, inspections or maintenance, repair and overhaul. This leads to excessive costs for customers and blocks specialists due to high amounts of manual work.
Annotation
Annotations take care about storing meta-data at data-points. This data can also be called information.
When processing objects, annotations allow correlations between information-sets for further processing. In the last years 2D objects, such as images, were processed; use-cases are addressing:
- Identification of objects
e.g. Amazon App for shopping of identified objects. - Dictionary Searches
e.g. for identification of flowers and illnesses. - Simultanious language processes
Convert languages into known ones, e.g. Google Lens in Japan.
Modern systems, such as cars or drones, require 3D objects, for identifying objects and these handle accurately for crash-control, lane assistance, auto-parking or autonomous driving scenarios. Other systems support agriculture, such as FarmBot. Therefore, Annotations are the key for proper processing of these data.
For annotating data, sensors are red, data is analyzed and an enhanced output is generated:
Validate
Deepen AI is one of the few tools, focusing on verified results. That’s due to the impact objects, such as cars, have. In San Francisco, cruise cars are already driving autonomously at night. The offer is planned for extension to day-time in 2023.
As of now, due to customer-insurance, a public offer is not yet available. If you would like to drive, you need to have someone – having singed a risk-paper – for taking a ride. If you’re interested how that might looks like: don’t hesitate contacting me, as I do have material available as well.