Hello, I am working on a project where we are doing object detection with respect to a world frame. In our case it is user defined and we are aiming to place objects at specific points in our world frame or “graph.” Using the built in object detection from the zed sdk we are only able to reliably detect people. Meaning that when a person is in frame it is always detected as a person and given an ID. If we try to use a different object for example a coffee mug, an apple, etc it is never reliably detected. Is this potentially something that we are doing wrong, or does it have something to do with the way the built in model was trained?
We are looking at trying a different model so that we can better detect these objects as we don’t want to be limited to detecting people. Ideally we are looking to detect more natural objects but at the moment we would like to be able to detect everyday objects and not just people.