Convert SVO to rosbag

,

Dear support,
We would like to record stereo images from ZED 1 camera along other sensors (as IMU data, GPS data, wheel odometry etc.), to evaluate different SLAM algorithms offline. We know that SVO is better than rosbag for stereo image recording but it does not support the recording of other sensors. Therefore, we would like to record a SVO for image coming from ZED and a rosbag for the rest of the sensors, to finally create a rosbag with all together.

The question that comes out then is:
if we record the SVO and a rosbag with the other sensors, How can we include the stereo images (inside the SVO) in to the rosbag (recorded with the rest of the sensors)?
Important: In order to put the stereo images in the rosbag properly we would need to extract from the SVO with their timestamps.

Thanks in advance,
Taihú

Hi @taihu
you can transform an SVO into a rosbag simply by starting a ZED Node using the SVO as input and then starting rosbag recording for the selected topics:
https://www.stereolabs.com/docs/ros/#launching-with-recorded-svo-video

Hi @Myzhar ,
I am recording an SVO file using this zed-examples/svo recording/recording at master · stereolabs/zed-examples · GitHub. When I launch the SVO file as you suggested, I get the rate of the IMU as 100, however I need it to be the max which is 400 (zed 2), I am looking into the other tutorials, and I can not find how to properly control the IMU rate. Would also be great to control other settings for the pose. Please let me know.

Hi @kha83263,
in the SVO you cannot record IMU samples at the maximum rate, but only the sample synchronized with the relative frame is saved.
To record IMU data at the maximum rate you must write your own separated logger or use directly rosbag save capabilities.

Thank you for the quick reply @Myzhar . I was trying with the rosbag, however, I need to store the images as well for visual inertial odometery testing. 1 min of raw data at 30fps, of HD720p is around 13GB. The compressed images are around 250MB which is amazing, however I am not sure how such high compression affects visual odometery techniques. If you have any suggestions, I would be really greateful. Thanks again.