Dynamically adjusting crop to input resolution

Accelerated Computing Intelligent Video Analytics DeepStream SDK
4

Thanks for your response. I’ve tested this out and decided to update, in case anyone else faces the same issue.

I followed official Gstreamer docs to make a probe and ended up with the following code:

// Registering probe
GstPad *source_srcpad = gst_element_get_static_pad (source_bin, "src");
gst_pad_add_probe(  source_srcpad, 
                        GST_PAD_PROBE_TYPE_BUFFER, 
                        (GstPadProbeCallback)test_probe, 
                        (gpointer)nvvidconv_crop, 
                        NULL);

// Probe callback
GstPadProbeReturn CounterPipeline::test_probe(GstPad *pad, GstPadProbeInfo *info, gpointer user_data)
{
    /**
     * Callback for source pad buffer probe.
    */
    GstCaps *current_caps = gst_pad_get_current_caps(pad);
    GstStructure *caps_structure = gst_caps_get_structure(current_caps, 0);
    gint height, width;
    GstElement *nvvidconv_crop = (GstElement *)user_data;

    if (gst_structure_get_int(caps_structure, "width", &width) && gst_structure_get_int(caps_structure, "height", &height))
    {
        if (width != 1920 || height != 1080) 
        {
            g_print("Resolution changed from %dx%d to %dx%d\n", 1920, 1080, width, height);
            std::string scr_crop = "100:100:200:200";
            g_object_set(G_OBJECT(nvvidconv_crop), "src-crop", scr_crop.c_str(), NULL);
        }
    }

Probing resolution worked as expected. Surprisingly, dynamically setting nvvideoconvert’s property worked as well!

Hope it helps someone in the future.