**• Hardware Platform (Jetson / GPU)**Jetson
• DeepStream Version6.1.1
**• JetPack Version (valid for Jetson only)**5.0.1
Dear all,I am trying to add my custom function to deepstream-app sample,the folder is like ：
And no long time before I have raised some topics, and according to some advices,I add my custom funtion to
gie_processing_done_buf_prob which is a part of deepstream_app.c,the whole code snippet is:
static GstPadProbeReturn
gie_processing_done_buf_prob (GstPad * pad, GstPadProbeInfo * info,
gpointer u_data)
{
GstBuffer *buf = (GstBuffer *) info->data;
NvDsInstanceBin *bin = (NvDsInstanceBin *) u_data;
guint index = bin->index;
AppCtx *appCtx = bin->appCtx;
//extra
NvDsBatchMeta *batch_meta = gst_buffer_get_nvds_batch_meta(buf);
// Get original raw data
GstMapInfo in_map_info;
if (!gst_buffer_map (buf, &in_map_info, GST_MAP_READ)) {
g_print ("Error: Failed to map gst buffer\n");
gst_buffer_unmap (buf, &in_map_info);
return GST_PAD_PROBE_OK;
}
NvBufSurface *surface = (NvBufSurface *)in_map_info.data;
for (NvDsMetaList * l_frame = batch_meta->frame_meta_list; l_frame != NULL;
l_frame = l_frame->next) {
NvDsFrameMeta *frame_meta = l_frame->data;
//TODO for cuda device memory we need to use cudamemcpy
NvBufSurfaceMap (surface, -1, -1, NVBUF_MAP_READ);
/* Cache the mapped data for CPU access */
NvBufSurfaceSyncForCpu (surface, 0, 0); //will do nothing for unified memory type on dGPU
guint height = surface->surfaceList[frame_meta->batch_id].height;
guint width = surface->surfaceList[frame_meta->batch_id].width;
float angle;
edgedetection(height,width,surface,frame_meta,& angle);
}
if (gst_buffer_is_writable (buf))
process_buffer (buf, appCtx, index);
return GST_PAD_PROBE_OK;
}
And the edgedetection is my custom funtion:
//custom function
void edgedetection(guint height,guint width,NvBufSurface *surface,NvDsFrameMeta *frame_meta,float * angle){
Mat nv12_mat = Mat(height*3/2, width, CV_8UC1, surface->surfaceList[frame_meta->batch_id].mappedAddr.addr[0],
surface->surfaceList[frame_meta->batch_id].pitch);
//Convert nv12 to RGBA to apply algo based on RGBA
Mat rgba_mat;
cvtColor(nv12_mat, rgba_mat, COLOR_YUV2BGRA_NV12);
for (NvDsMetaList * l_obj = frame_meta->obj_meta_list; l_obj != NULL;
l_obj = l_obj->next) {
NvDsObjectMeta *obj = (NvDsObjectMeta *) l_obj->data;
Rect bbox(obj->rect_params.left, obj->rect_params.top,
obj->rect_params.width, obj->rect_params.height);
Mat roi = rgba_mat(bbox);
float angle;
Ellipse_feature_extraction(roi,angle);
}
}
Ellipse_feature_extraction is a funtion to apply edgedetection to get the angle of each detected object:
void Ellipse_feature_extraction(const Mat& img,float& angle) {
if (img.empty()) {
printf("Received an empty image\n");
return;
}
Mat imgray;
cvtColor(img, imgray, COLOR_BGRA2GRAY);
Mat edges;
Canny(imgray, edges, 500, 100);
Mat thresh;
threshold(edges, thresh, 127, 255, THRESH_BINARY);
std::vector<std::vector<Point>> contours;
std::vector<Vec4i> hierarchy;
findContours(thresh, contours, hierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE);
for (const auto& cnt : contours) {
if (cnt.size()>100) {
RotatedRect rect = minAreaRect(cnt);
Point2f box[4];
rect.points(box);
angle=rect.angle;
}
}
}
The logic is that I need to get the raw image data of each frame,namely the rgba_mat in my custom funtion.And in every frame loop,the frame image will be managed in edgedetection function to access every object.After that,I will get the ROI image of each object and apply edge detection using Ellipse_feature_extraction.After that,I may get the angle of each detected object.
Then I sudo make install in the terminal,everything is fine with no errors.But when I try:
deepstream-app -c deepstream_app_config.txt
to start the pipeline.The pipeline seems to crashes with only one frame appear then the pipeline ends:
I must express my sincere thanks if you can read the complete text,I just wonder why the pipeline just ends suddenly,and there must be something I’m not thinking about.