I want to know how streammux resize input and how enable-padding works in streamxux. Where can I find it? Thanks.
Please refer to our Guide: https://docs.nvidia.com/metropolis/deepstream/dev-guide/text/DS_plugin_gst-nvstreammux.html#id2
@yuweiw
Thanks. It is a guide to config. But I want to investigate source code of streamxux. Where is it?
This plugin is not open source. If you have any issues when use that, you can file a new topic.
@yuweiw
Thanks for response.
I have some questions:
- if I set
enable-padding=1for streamux, so padding is done for top, left, bottom, right. Is this right? - if I set
enable-padding=0for streamux, so padding is done for top, left. Is this right?
In 2 above cases, firstly image size is scaled to streamux size by using max dimension. Is this right?
Is there any patch to save frames right after streamux?
No. The padding is for Maintains aspect ratio. It scales first, then padding.
Basically, you can get the raw data from any of the plugin by adding the probe function to it. You can refer to the patch below:
deepstream_image_decode_app_dump_data.patch (7.8 KB)
Yes, padding for maintain aspect ratio.
Basically, you can get the raw data from any of the plugin by adding the probe function to it. You can refer to the patch below:
I will try to dump data. Thanks.
@yuweiw
I have one more question.
Algorithm to scale and padding image in streamux is same with algorithm to scale and padding image in nvinfer before inference (set maintain_aspect_ratio=1 and set padding=1)?
Thanks.
Yes. And the nvinfer is open source. You can refer to that.
Thanks.
@yuweiw
could yiu tell me please what kind of interpolation is user for resizing with enable padding in streamumux? Is this bilinear?
Please refer to our Guide : scaling-filter.
This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.