I’ve noticed that in turotial 9, once the robot (gazebo) sees the object, it doesn’t immediately set “object_found” to true yet. It actually gets closer and cenetered with the object, then sets it to true. I would like to use this same program, but implement my own tracking method (Tensorflow) instead. Could you point out exactly which parts of your tracking code I can replace with my own tracking to get this working? Thanks.
Hello geasrmotion789,
Object detection is preformed by find_object_3d node from find_object_2d package.
You can find its configuration in file find_object_3d.launch.
Node find_object_3d subscribe image from camera and publish detected objects as std_msgs/Float32MultiArray along with tf frame for object position.
You will need to replace this node with your own implementation.
Regards,
Łukasz