Deepstream preprocess element output seems to be the same as input

**• Hardware Platform: GPU
**• DeepStream Version: 6.4
**• NVIDIA GPU Driver Version: 535.104.12
**• Issue Type: question/bug
**• How to reproduce the issue ?

When analyzing preprocess element output, using a probe function on the preprocess src pad - it seems the frame size remains the same as the input - or, if we are looking in the wrong place, where should we see the reshaped frame?

  • Based on DS python apps/deepstream-preprocess-test
  • Preprocess width and hight config remains as default: 960, 544
  • running command:

python3 -i file:///opt/nvidia/deepstream/deepstream/samples/streams/sample_720p.mp4

in the preprocess src probe function:

data_type, shape, strides, dataptr, size = pyds.get_nvds_buf_surface_gpu(hash(gst_buffer), frame_meta.batch_id)

> shape=(1080, 1920, 4)
Although we expect it to be 544x960x4 . is that right?

*in order to use get_nvds_buf_surface_gpu() we must to add conversion to RGBA to this pipeline:

nvvidconv0 = Gst.ElementFactory.make(“nvvideoconvert”, “convertor0”)
filter0 = Gst.ElementFactory.make(“capsfilter”, “filter0”)
caps1 = Gst.Caps.from_string(“video/x-raw(memory:NVMM), format=RGBA”)
filter0.set_property(“caps”, caps1)

1 Like

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

We do not transform the original image data in the Gstbuffer. It’s scaled in the preprocess plugin. If you want to get the scaled image, you can use the dump_rois in the gstnvdspreprocess.cpp.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.