So we wanted to share a bit of the background on object detection, our experience with it, and why we made DepthAI to allow object localization.
Long-story short, object detection seems like all you would need - until you start working with the physical world. Then you quickly realize it falls short, unless you can somehow work in a constrained, pre-calibrated environment.
That’s where DepthAI comes in. Boom. Object Localization - showing you what the object is, where it is, in physical coordinates (i.e. that apple is 4 feet in front of me, 3 inches above me, and 5 inches to the left) and all in real time. And oh by the way, the neural model says that it’s ripe and ready to pick. "Robotic arm, grab it at those coordinates."
Or, you can tell where the car is relative to the camera, what the license plate is, and how fast the car is going (and it’s full 3D trajectory, without having to pre-calibrate the scene, etc.)
Brandon & the Luxonis Team