Generated GRIP code is wildly different

After getting the vision code set up so that the RGB filtered image is output to the SmartDashboard, we have noticed that the thresholding between the robot and computer is vastly different. Has anyone else had this issue? Is there a fix for this?

What kind of camera are you using? How is GRIP connected to the camera?

Axis camera through a CameraServer.

Not trying to bump, but we have tried changing the camera resolution as well and no joy.