When you're looking at a tank full of fish at a public aquarium, it can be difficult to figure out which ones are which species. The AI Aquarium is designed to help, by overlaying information that lines up with the individual fish in question.
Created by Taiwan's Industrial Technology Research Institute (ITRI), the AI Aquarium was recently the recipient of a CES Innovation Award. Here's how it works …
An outward-facing depth-sensing 3D camera located above the aquarium tracks the viewer's eyes, determining where they're looking within the tank. Another camera pointing into the aquarium tracks the positions of all the fish. Real-time video from that second camera is analyzed by an object recognition algorithm, which visually matches each fish to photos of known species in a computer database.
By comparing the viewer's line of sight to the positions of the fish, the AI Aquarium can thus determine not only which individual fish the person is currently observing, but also what species that fish belongs to. This information is displayed on a transparent microLED screen overlying the front glass of the aquarium, in such a way that the text is overlaid directly beside the person's view of the fish.
If viewers want more information than just the fish's name, they use simple hand gestures to prompt the system to display more details.
The AI Aquarium system is already in use at Taiwan's National Museum of Marine Science and Technology. It can be seen in use, in the video below.
Source: ITRI