[setting 1]
batchsize = 2
infervalve’s drop = FALSE
menuUIvalve’s drop = TRUE
tiler = 1rows, 2column → looks good
[setting 2]
batchsize= 2
infervalve’s drop = TRUE
menuUIvalve’s drop = FALSE
->throw error in (ERROR: batch size not 1)**
why throw this error?
help me.
thanks…
+++++++++++Add
I couldn’t find nvoverlaysink’s info in Nvidia Deepstream Details Doc
but i think nvoverlaysink’s input must contained only Single frame buffer
and think The reason of the Deepstream’s error is because the Dsexample’s output is ‘N batched data’
i want to know how to processing(N batched buffer → dsexample’s transform_ip Process → single frame buffer)
Like nvmultistreamtiler or compositor
what I want to make is change over from streammux’s src data(N Batched data) to single batched data(composited) after processing my customplugin’s transform function
like i wrote before i think overlaysink has to contain only one batched data
i found function called “transform” in GstBaseTransform API Reference doc, i think if use it i can make i want
i tried.(referred to topic of similar problem)
my transform code is below, but something seems to be wrong.
I found some functions from gstbasetransform API for modifying output buffer, their names are “prepare_output_buffer” and “transform”
I added a GstBuffer to struct of _GstMenuDraw (The name is different from Dsexample and the structure is the same. ), and i made new GstBuffer on gst_menudraw_start().
code is below (name of the added gstbuffer is “outBuff”)
// Create host memory for storing converted/scaled interleaved RGB data
CHECK_CUDA_STATUS (cudaMallocHost (&menudraw->host_rgb_buf,
menudraw->processing_width * menudraw->processing_height *
RGBA_BYTES_PER_PIXEL), “Could not allocate cuda host buffer”);
GST_DEBUG_OBJECT (menudraw, “allocated cuda buffer %p \n”,
menudraw->host_rgb_buf);
// CV Mat containing interleaved RGB data. This call does not allocate memory.
// It uses host_rgb_buf as data.
menudraw->cvmat =
new cv::Mat (menudraw->processing_height, menudraw->processing_width,
CV_8UC3, menudraw->host_rgb_buf,
menudraw->processing_width * RGBA_BYTES_PER_PIXEL);