Possible to remove frame input of SVO recording


My project would like to reduce the size of the SVO recording by some filters on input image pair while doing recording. To save recording space, I want to apply some filters before recording into SVO file.

Is there any ways to do this?

Thank you.

Hi @linh.nguyen,

Can you provide more information on what you mean by “filters” ?

Currently there are different compression methods to choose from in our API, with lossy methods like H265 and H264 to reduce disk space usage.


The “filter” here means to choose to keep the input frame or drop it from the recording. Likes a object detection to check if the frame is interested or not, if not, just not include the frame into recording.

I know we have lossy methods, but I’m trying to use even less storage space.

Hi again,

Thank you for clarifying. There are no such mechanisms as the recording is performed internally in the grab() method.

What can be done is to call the grab only at specific punctual moments, and by doing this you will have an SVO with sparse images in time. However, you would like to process the image in order to choose whether to apply the filter or not.

What I would suggest in this case is to create a new recording for event-based situations like this one. In the API you can choose to start or stop the recording whenever an object is detected.

1 Like