Recording with a D415 upside down

[sming2]

Hi,

In order to perform a proof of concept, we have fixed two D415 from the ceiling using a camera clip. The problem is that the camera is upside down, which means that they are recording an inverted image. Both cameras are connected to a Win10 Intel Core i5 with GPU Nvidia GTX 1060 6B. For now, we are recording and processing the resulting bags, meaning we are not processing anything in real-time.

We have experienced some issues regarding dropping frames using the pipeline of the realsense. However, we could avoid missing information, by extracting the frames from the bag file to an HDF5. Since both colour and depth frames were not well sync/aligned due to the different rostopic publishing frequencies, we had to use the timestamps of each sensor to align the frames -the header timestamp of the rostopic-. We used to colour frames to fill the depth frames. When we passed the resulting frames to .avi, we saw that the depth frames had an x and y offset, meaning both channels not only were not perfectly well aligned in time which is more or less acceptable, but also in space. We wonder if the reason might be the fact that we are recording with the camera rotated 180. It is supposed that the D400 series, is already calibrated, but would it be necessary any additional or special calibration? Thanks. 

[MartyG · Accepted solution]

I'm not aware of the orientation of the 400 Series cameras making a difference to calibration or whether frames are dropped. For example, a drone might mount the camera facing downward instead of straight ahead in order to capture the ground passing under it. Also, frames can be dropped from a bag even when recording with the camera the normal way up.

There is a good discussion about dropped frames in the link below.

https://github.com/IntelRealSense/librealsense/issues/2535