I wonder what does infer-dims do in deepstream.
In the manual, it says Binding dimesnsions to set on the image input layer.(ex: infer-dims=3;224;224)
Does this mean that it would change the incoming stream to 224x224 frame and inference from there?
Also, the 3 in the example is supposed to mean channels.
What does channels mean in this example? I have only seen 3 for all cases.
It will convert and scale first based on network input, here by network width and height 224*224, you can see the FAQ about how nvinfer works. DeepStream SDK FAQ comment 1#
and for RGB or BGR format input, the channel is 3, for gray input, the channel is 1.
Hi @Amycao
Thanks for the reply.
Can you let me know what happens when 224 * 224 image comes in, (I guess it will pass itself as-is.
and when 448 * 448 comes in, and 896 * 896 comes in?
Thanks