I want to get pic data by using gstreamer-1.0 C interface,
but when run this code,my app crashed with “bus eeror”.
gst_element_factory_make ("nvjpegdec", "myjpegdec");
Does nvjpegdec support this way?
Hi,
We don’t observe the issue. Please refer to attachment
test2.zip (1.2 KB)
Thanks,
it works by raplacing
gst_element_factory_make ("nvjpegdec", "myjpegdec");
with
gst_parse_launch(launch_string.c_str(), &error);
But it will cost 100ms to decode a jpeg which resolution is 4096 * 2160.
Is the speed normal?
I think it will decode more fast.
It looks OK for decoding 4K jpeg. You may run jetson_clocks.sh to get better performance.
For 4Kp30 decoding, we suggest you use H264/H265.
I used gstreamer-1.0 in cmake file as below:
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -pthread")
cmake_minimum_required(VERSION 3.0 FATAL_ERROR)
project(camera_agent CXX)
find_package(PkgConfig REQUIRED)
find_package(OpenCV REQUIRED)
include_directories(
/usr/local/include
${GLIB_INCLUDE_DIRS}
)
link_directories(
${GLIB_LIBRARY_DIRS}
)
set(GSTREAMER_MINIMUM_VERSION 1.0.5)
pkg_check_modules(GST1_TEST gstreamer-1.0)
if ( GST1_TEST_FOUND AND NOT ${GST1_TEST_VERSION} VERSION_LESS ${GSTREAMER_MINIMUM_VERSION} )
pkg_check_modules(GSTREAMER REQUIRED gstreamer-1.0)
pkg_check_modules(GSTREAMER-APP REQUIRED gstreamer-app-1.0)
pkg_check_modules(GSTREAMER-AUDIO REQUIRED gstreamer-audio-1.0)
pkg_check_modules(GSTREAMER-PBUTILS REQUIRED gstreamer-pbutils-1.0)
pkg_check_modules(GSTREAMER-FFT REQUIRED gstreamer-fft-1.0)
add_definitions(-DGST_API_VERSION_1=1)
else()
# fallback to gstreamer-1.0
unset(GSTREAMER_MINIMUM_VERSION)
pkg_check_modules(GSTREAMER REQUIRED gstreamer-1.0)
pkg_check_modules(GSTREAMER-APP REQUIRED gstreamer-app-1.0)
pkg_check_modules(GSTREAMER-AUDIO REQUIRED gstreamer-audio-1.0)
pkg_check_modules(GSTREAMER-FFT REQUIRED gstreamer-fft-1.0)
set_source_files_properties(WebKitWebAudioSourceGStreamer.cpp PROPERTIES COMPILE_DEFINITIONS "GLIB_DISABLE_DEPRECATION_WARNINGS=1")
endif()
include_directories(
${GSTREAMER_INCLUDE_DIRS}
${GSTREAMER-APP_INCLUDE_DIRS}
${GSTREAMER-AUDIO_INCLUDE_DIRS}
${GSTREAMER-PBUTILS_INCLUDE_DIRS}
${GSTREAMER-FFT_INCLUDE_DIRS}
${CMAKE_CURRENT_SOURCE_DIR}
include
)
link_directories(
${GSTREAMER_LIBRARY_DIRS}
${GSTREAMER-APP_LIBRARY_DIRS}
${GSTREAMER-AUDIO_LIBRARY_DIRS}
${GSTREAMER-PBUTILS_LIBRARY_DIRS}
${GSTREAMER-FFT_LIBRARY_DIRS}
/usr/lib/
)
set(inputtest_LIBRARIES
${GSTREAMER_LIBRARIES}
${GSTREAMER-APP_LIBRARIES}
${GSTREAMER-AUDIO_LIBRARIES}
${GSTREAMER-PBUTILS_LIBRARIES}
${GSTREAMER-FFT_LIBRARIES}
pthread
${GLIB_LIBRARIES}
${GLIB_GIO_LIBRARIES}
${GLIB_GOBJECT_LIBRARIES}
)
add_executable(
camera_agent
src/camera_agent.cpp
)
target_link_libraries(
camera_agent
v4l2
uvc_camera
zmq_util
agent_manager
thread_manager
cppzmq
opencv_camera
memory_manager
config_manager
${OpenCV_LIBS}
glog
${inputtest_LIBRARIES}
)
# set the output directory and output file name
set_target_properties(camera_agent
PROPERTIES
OUTPUT_NAME "camera_agent"
RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/${CMAKE_INSTALL_BINDIR}/")
It will print “core dumped”,when running this code:
gst_element_factory_make ("nvjpegdec", "myjpegdec")
However,it will run normally by using Makefile as below:
TARGET=sample
CC=g++
SRC=$(wildcard *.cpp)
OBJ=$(SRC:.cpp=.o)
CXXFLAGS=-g
CXXFLAGS += `pkg-config --cflags gstreamer-1.0 gstreamer-app-1.0 glib-2.0`
LDFLAGS += `pkg-config --libs gstreamer-1.0 gstreamer-app-1.0 glib-2.0` -lpthread
all:$(TARGET)
$(TARGET):$(OBJ)
$(CC) $(CFLAGS) -o $@ $^ $(LDFLAGS)
clean:
rm -f $(TARGET) $(OBJ)
I want to use in cmake file.Would you please give me your hand ?
try to use gst_parse_launch(launch_string.c_str(), &error) instead of gst_element_factory_make (“nvjpegdec”, “myjpegdec”) in your code.
Thanks a lot.
It does’t work.
I have the same problem. I get:
[gstreamer] initialized gstreamer, version 1.14.5.0
Bus error (core dumped)
when I run
std::string launchStr = "v4l2src device=/dev/video0 ! image/jpeg, width=640, height=480, framerate=30/1 ! nvjpegdec ! video/x-raw ! nvvidconv ! 'video/x-raw(memory:NVMM), format=NV12' ! videoconvert ! appsink name=mysink";
gst_parse_launch(launchStr.c_str(), &error)
Does anyone know how to resolve this problem by any chance? nvjpegdec works perfectly when I launch the pipeline above through the command line (gst-launch-1.0).
Hi,
Please check the following post:
https://devtalk.nvidia.com/default/topic/1070087/jetson-tx2/bus-error-with-gstreamer-and-opencv-/post/5422014/#5422014
nvjpegdec is not compatible with OpenCV. Please use jpegdec plugin.
@DaneLLL I used only the C++ API for Gstreamer and didn’t actually use OpenCV (gst_parse_launch is included from <gst/gst.h>, I believe). I also added proper include paths into my CMakeLists.txt file for GStreamer.
Hi,
Have you tried the sample in #2?
@DaneLLL Please could you share the Makefile/CMakeLists.txt/g++ command that you used to compile the file and include the libraries? I am getting an error when compiling the file with my project’s CMakeLists.txt:
//usr/lib/aarch64-linux-gnu/libgobject-2.0.so.0: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
Update: was able to compile and test the source file in #2 using @wangxin112’s CMakeLists.txt - it works! Looks like it’s the problem with dependencies in the first project. Will investigate further.
Hi,
For your reference:
nvidia@nvidia-desktop:~$ g++ test2.cpp -o test2 $(pkg-config --cflags --libs gstreamer-app-1.0)
nvidia@nvidia-desktop:~$ ./test2
Using launch string: filesrc location=a.jpg ! nvjpegdec ! video/x-raw ! appsink name=mysink
2976x2976, map.size = 13284864
app sink receive eos
You were absolutely right! The error was due to having cv::imread in one of the source files (even if I don’t call the function in my code). If I replace nvjpegdec with jpegdec or if I delete the cv::imread call, then the error goes away.
Do you know why would that be the case by any chance? libjpeg gets linked instead of libnvjpeg, if I define cv::imread in my source files?
Hi,
You may give it a try. Or use libjpeg while building application with OpenCV.
Currently libnvjpeg does not work with OpenCV.