How do I get a picture with a rectangular frame?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

I need to get a picture with a rectangular box, and now I get the picture from frame_meta, but there is no rectangular box. I don’t know why, I hope you can help me.
This is my code:

for (l_frame = batch_meta->frame_meta_list; l_frame != NULL;
    l_frame = l_frame->next) {
      NvDsFrameMeta *frame_meta = (NvDsFrameMeta *) (l_frame->data);
      gint frame_width = (gint)nvsurface->surfaceList[frame_meta->batch_id].width;
      gint frame_height = (gint)nvsurface->surfaceList[frame_meta->batch_id].height;
      void *frame_data = nvsurface->surfaceList[frame_meta->batch_id].mappedAddr.addr[0];
      size_t frame_step = nvsurface->surfaceList[frame_meta->batch_id].pitch;
      in_mat = cv::Mat(frame_height,frame_width,CV_8UC4,frame_data,frame_step);
      out_mat = cv::Mat(frame_height,frame_width,CV_8UC3);
      cv::cvtColor(in_mat,out_mat,CV_RGBA2BGR);
2

HI
you want to draw detected objects, is this what you mean, if yes, the info you need is in NvDsObjectMeta,
refer to this, NVIDIA Metropolis Documentation,
you need to iterate through object meta list within frame, fill struct GstNvOsd,
and draw it over frame buffer,
refer to: int nvll_osd_draw_rectangles(void *nvosd_ctx, NvOSD_FrameRectParams *frame_rect_params);
which from sources/includes/nvll_osd_api.h

3

Thanks amycao, I will try this.

4

Hello，amycao.
The picture I am getting now is obtained from frame_meta using NvBufSurface.
I found int nvll_osd_draw_rectangles (void * nvosd_ctx, NvOSD_FrameRectParams * frame_rect_params) based on your reply, but I do n’t know how to use it, can you provide me a copy Its code? If I could, I would be grateful.
Best wishes to you！

5

oh, this API supposed to draw buffer blended from original buffer and NvOSD_FrameRectParams on the screen, but you want to save this blended buffer to one local file, Does using sink type 3 File meet your requirements?