For the ATAP team:
My Yellowstone device appears to have an IR filter on the fish-eye camera lens. I have tried the Tango in a dark room flooded with IR light. The color camera shows the room lit up (as grayscale), and the fisheye shows nothing (just a black image). The Tango Service does not work in this setting.
How critical is the fish-eye for visual odometry? I'm wondering if removing the IR filter from the fish-eye lens would enable the Tango service to give correct positional/orientation readings. But it's an expensive operation to be performed on my Tango, so I wanted to ask the question before attempting it.
FYI, the Tango Service gives status messages like "FishEye underexposed" in this setting. And perhaps the Tango Service has stopped trying to perform visual odometry in this state (even though the RGB-IR camera can see the room as greyscale). So to trick out the Tango Service, I used a small penlight & shined it into the fish-eye camera. This got rid of the underexposure status, but it did not enable correct orientation & positioning. The Tango was still quite lost. As near as I can tell, either no visual odometry is being done with the RGB-IR camera (so the fisheye does it all)...or camera frames lit by IR are not acceptable for visual odometry (perhaps color is needed for natural feature recognition???).
Can the Tango team provide any insight in how visual odometry could be performed in a dark room by IR light alone? Or is this even possible?