This example shows how to use Intel RealSense cameras with existing Deep Neural Network algorithms. The demo is derived from MobileNet Single-Shot Detector example provided with
opencv. We modify it to work with Intel RealSense cameras and take advantage of depth data (in a very basic way).
The demo will load existing Caffe model (see another tutorial here) and use it to classify objects within the RGB image. Once object is detected, the demo will calculate approximate distance to the object using the depth data:
Unlike the other samples, this demo requires access to the exact depth values. We generate a matrix of floating point values (in meters) using the following helper function:
auto depth_mat = depth_frame_to_meters(pipe, depth_frame);