Overall flow about deepstream architecture

In this graph architecture,

  1. Why image processing and DNN boxes are two?

  2. Can anyone explain the reason why each step use CPU/GPU or NVDEC such like that?
    I want to understand the overall flow about deepstream architecture but it is harsh for me because I am the beginner of deepstream.

  3. How each plugin is connected to each step, and if you attached the code for them is better for me to understand. I am eagerly want to understand it, thank you for your further answer.

Why image processing and DNN boxes are two?

–> means batch processing

Can anyone explain the reason why each step use CPU/GPU or NVDEC such like that?

–> means the component running under which hardware

I want to understand the overall flow about deepstream architecture but it is harsh for me because I am the beginner of deepstream.

–> the picture you posted shown deepstream overall architecture, for specific component, source, streammux, nvinfer, osd, tiler, etc. you can get details here,
https://docs.nvidia.com/metropolis/deepstream/plugin-manual/index.html#page/DeepStream_Plugin_Manual%2Fdeepstream_plugin_details.02.01.html,

How each plugin is connected to each step, and if you attached the code for them is better for me to understand. I am eagerly want to understand it, thank you for your further answer.

–> we have kinds samples, each sample demonstrates how to use deepstream SDK elements for different usercase, location, sources/apps/sample_apps/, let’s take test1 as example about how the elements connected, file-source -> h264-parser -> nvh264-decoder -> nvinfer -> nvvidconv -> nvosd -> video-renderer