Nvds_obj_enc_process SIGSEGV

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform: Jetson AGX Orin)
• DeepStream Version: 6.1
• JetPack Version: 5.0.2.
• TensorRT Version: 8.4.1
• Issue Type: Question

I am trying to run the deepstream-image-meta-test sample found here: /opt/nvidia/deepstream/deepstream-6.1/sources/apps/sample_apps/deepstream-image-meta-test/

This sample app out-of-the-box gives a segmentation fault in nvds_obj_enc_process while trying to save images. I modified the pgie_src_pad_buffer_probe to save full frames instead:

static GstPadProbeReturn
pgie_src_pad_buffer_probe (GstPad * pad, GstPadProbeInfo * info, gpointer ctx)
{
  GstBuffer *buf = (GstBuffer *) info->data;
  GstMapInfo inmap = GST_MAP_INFO_INIT;
  if (!gst_buffer_map (buf, &inmap, GST_MAP_READ)) {
    GST_ERROR ("input buffer mapinfo failed");
    return GST_FLOW_ERROR;
  }
  NvBufSurface *ip_surf = (NvBufSurface *) inmap.data;
  gst_buffer_unmap (buf, &inmap);

  NvDsObjectMeta *obj_meta = NULL;
  guint vehicle_count = 0;
  guint person_count = 0;
  NvDsMetaList *l_frame = NULL;
  NvDsMetaList *l_obj = NULL;
  NvDsBatchMeta *batch_meta = gst_buffer_get_nvds_batch_meta (buf);

  for (l_frame = batch_meta->frame_meta_list; l_frame != NULL;
      l_frame = l_frame->next) {
    NvDsFrameMeta *frame_meta = (NvDsFrameMeta *) (l_frame->data);

    // Create a new NvDsObjectMeta for the entire frame
    NvDsObjectMeta full_frame_meta = {0};
    full_frame_meta.rect_params.left = 0;
    full_frame_meta.rect_params.top = 0;
    full_frame_meta.rect_params.width = frame_meta->source_frame_width;
    full_frame_meta.rect_params.height = frame_meta->source_frame_height;
    full_frame_meta.class_id = PGIE_CLASS_ID_VEHICLE; // or any other class ID

    NvDsObjEncUsrArgs userData = { 0 };
    /* To be set by user */
    userData.saveImg = save_img;
    userData.attachUsrMeta = attach_user_meta;
    /* Set if Image scaling Required */
    userData.scaleImg = FALSE;
    userData.scaledWidth = 0;
    userData.scaledHeight = 0;
    /* Preset */
    userData.objNum = 1;
    /* Quality */
    userData.quality = 80;
    /*Main Function Call */
    nvds_obj_enc_process (ctx, &userData, ip_surf, &full_frame_meta, frame_meta);
  }
  nvds_obj_enc_finish (ctx);
  return GST_PAD_PROBE_OK;
}

Now, the first full frame is saved, but the segmentation fault still occurs in the call to nvds_obj_enc_finish. Analyzing the stack trace, the issue is with nvds_acquire_user_meta_from_pool which is internally called by the API.

Here is another forum post detailing my exact issue: Calling nvds_obj_enc_finish SIGSEGV - #8 by doug4350

Is there any working demo or correct resource usage for encoding the entire image? Is saving the entire frame even supported?

2

Hi

You need to allocate an object meta using nvds_acquire_obj_meta_from_pool when creating full_frame_meta instead of manually instantiating it.

Regards,
Allan Navarro

Embedded SW Engineer at RidgeRun

Contact us: support@ridgerun.com
Developers wiki: https://developer.ridgerun.com/
Website: www.ridgerun.com

3

Hi, thank you for responding so soon. I’ve modified the code to do that and still receive a seg fault in the same place:

static GstPadProbeReturn
pgie_src_pad_buffer_probe (GstPad * pad, GstPadProbeInfo * info, gpointer ctx)
{
  GstBuffer *buf = (GstBuffer *) info->data;
  GstMapInfo inmap = GST_MAP_INFO_INIT;
  if (!gst_buffer_map (buf, &inmap, GST_MAP_READ)) {
    GST_ERROR ("input buffer mapinfo failed");
    return GST_FLOW_ERROR;
  }
  NvBufSurface *ip_surf = (NvBufSurface *) inmap.data;
  gst_buffer_unmap (buf, &inmap);

  NvDsObjectMeta *obj_meta = NULL;
  guint vehicle_count = 0;
  guint person_count = 0;
  NvDsMetaList *l_frame = NULL;
  NvDsMetaList *l_obj = NULL;
  NvDsBatchMeta *batch_meta = gst_buffer_get_nvds_batch_meta (buf);

  for (l_frame = batch_meta->frame_meta_list; l_frame != NULL;
      l_frame = l_frame->next) {
    NvDsFrameMeta *frame_meta = (NvDsFrameMeta *) (l_frame->data);

    // Allocate a new NvDsObjectMeta for the entire frame
    NvDsObjectMeta *full_frame_meta = nvds_acquire_obj_meta_from_pool(batch_meta);
    if (!full_frame_meta) {
      GST_ERROR ("Failed to acquire object meta from pool");
      return GST_FLOW_ERROR;
    }
    full_frame_meta->rect_params.left = 0;
    full_frame_meta->rect_params.top = 0;
    full_frame_meta->rect_params.width = frame_meta->source_frame_width;
    full_frame_meta->rect_params.height = frame_meta->source_frame_height;
    full_frame_meta->class_id = PGIE_CLASS_ID_VEHICLE; // or any other class ID

    NvDsObjEncUsrArgs userData = { 0 };
    /* To be set by user */
    userData.saveImg = save_img;
    userData.attachUsrMeta = attach_user_meta;
    /* Set if Image scaling Required */
    userData.scaleImg = FALSE;
    userData.scaledWidth = 0;
    userData.scaledHeight = 0;
    /* Preset */
    userData.objNum = 1;
    /* Quality */
    userData.quality = 80;
    /*Main Function Call */
    nvds_obj_enc_process (ctx, &userData, ip_surf, full_frame_meta, frame_meta);
  }
  nvds_obj_enc_finish (ctx);
  return GST_PAD_PROBE_OK;
}
4

I think using nvds_obj_enc_process might not be the best idea to save the full frame, maybe using the GstBuffer might be better, let me check if I have some sample for that.

Regards,
Allan Navarro

Embedded SW Engineer at RidgeRun

Contact us: support@ridgerun.com
Developers wiki: https://developer.ridgerun.com/
Website: www.ridgerun.com

6

@shaunramj , could you upgrade your DeepStream to our latest version? You can just set the userData.isFrame=1 without changing any other code to save the full frame.

7

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

8

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.