For privacy reasons I need help censoring .bag files that were recorded in a hospital #13010
Replies: 2 comments 1 reply
-
|
Hi @antoniogrimalt There was an issue in recent RealSense Viewer versions where bag files recorded in an older SDK version could not be played back in the current version. Only bags recorded in the same Viewer version that they were played back in would work. This issue was apparently resolved in version 2.55.1. Have you had the opportunity to test your recorded ROS rosbags in 2.55.1 please? |
Beta Was this translation helpful? Give feedback.
-
|
Yes, I've tested every version of the Viewer executables that are linked in the releases. The original .bag files that were recorded with the Viewer all open correctly, in the last version and in the previous ones. The problem comes when I try to play a trimmed version created with rosbag. Only 2.48 and lower can play them correctly. |
Beta Was this translation helpful? Give feedback.
-
|
Checking the SDK release notes for 2.49, there were a couple of changes to how rosbags are handled.
|
Beta Was this translation helpful? Give feedback.
-
Hello RealSense community!
I am doing a project in Python 3 that uses .bag files previously recorded with the D435i and D405 cameras (firmware v05.13.00.50). These recordings were made in a hospital, so the faces of doctors and patients sometimes appear in the color and depth streams.
My intention when starting the project was to censor the recordings by applying blur to the faces, but I haven't yet found a way to do this using pyrealsense2. I don't know how I can make two pipelines work at the same time, one reading from a .bag and the other writing to a new one. In the documentation I've read that it's not possible to use the methods enable_device_from_file() and enable_record_to_file() in the same config. I've tried doing it in different ways and I always get the same error:
Something I would like to know, because perhaps I'm trying something that isn't possible, is:
Can a new .bag be recorded with pyrealsense2 if the source is an existing .bag and not a camera?
I've come across other examples in which they are able to apply post processing between capturing the images from a camera and recording them in a new .bag, but I haven't seen any functional examples of my case.
As an alternative, although it's a worse solution, I've chosen to use the rosbag Python API and cut the video into clips that don't include faces. The problem I have found with this is that, although I include all the topics and their initial messages as they appeared in the original file, the latest versions of the SDK don't allow me to load or play the file without giving me errors.
I've found that the last version of the SDK that can play .bag files created with rosbag is v2.48.0. From the v2.49.0 to the v2.53.1 the color and depth streams appear white throughout the recording, as can be seen in this comment. Starting with the v2.54.1 version, the file no longer loads and the following error appears in the console:
If I publish the dataset of the censored .bag files, I would like them to be compatible with the latest versions of the SDK.
As I have read in other comments, I understand that rosbag .bag files are not fully compatible with the SDK. Even so, I am curious to know why in previous versions these .bag files could be loaded and viewed.
Does anyone know the reason why these .bag files can no longer be opened by newer SDK versions?
This is the first time I've started a discussion on GitHub, so if I need to include more information or change something so you can help me, I'll do it as soon as possible.
Beta Was this translation helpful? Give feedback.
All reactions