Pill Recognition - The Technology Challenge

Why is it so difficult to “teach” a phone to automatically tell which medications it “sees” ?

Irody’s end-to-end computer vision system tailored for pill identification include controlled image captured, real time video processing, and image analysis. To meet rigorous performance requirements, Irody develops, evaluates, and applies computer vision technologies in these areas: real-time object tracking in a video stream, object recognition, color processing, shape recognition and high-level artificial intelligence.

We provide the brain

The phone’s camera is performing a function similar to the human eye: it captures the images. Irody’s proprietary algorithms act as the human brain: interpret what is in the image – i.e. which pill or pills are present in the image. The human brain is extremely good in performing visual tasks, primarily because close to 50% of the brain “hardware” is dedicated to these tasks. This is why humans perform vision related tasks so naturally and effortlessly. A computer, in the contrary, is more suitable for adding and multiplying numbers, not to perform high level visual interpretation tasks. Irody needed to develop cutting-edge technology for making a computer perform human-like vision tasks.

Using Artificial intelligence

The technology developed by Irody belongs to a family of “intelligent” programs such as programs that can understand spoken words – used by apps such as Apple’s Siri, or by automated call agents that ask you to say the reason for your call. Just like other AI-based solutions,  the performance is improving over time.