Segmentation fault when sending message through rabbitmq

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

I get Segmentation fault after sending three massage. I applied my model on deepstream_test4 and also, I doing post-processing on SGIE

PGIE Config:

[property]
gpu-id=0
process-mode=1

net-scale-factor=0.0039215697906911373
model-engine-file=//home/riotu/codes/models/Secondary_FaceDetect/fd_lpd.caffemodel_b1_gpu0_fp32.engine
labelfile-path=/home/riotu/codes/models/Secondary_FaceDetect/labels.txt

model-file=/home/riotu/codes/models/Secondary_FaceDetect/fd_lpd.caffemodel
proto-file=/home/riotu/codes/models/Secondary_FaceDetect/fd_lpd.prototxt

force-implicit-batch-dim=1
batch-size=1
network-mode=0

num-detected-classes=3
interval=2

gie-unique-id=2
#operate-on-gie-id=1

SGIE Config

[property]
gpu-id=0
process-mode=2

#net-scale-factor=0.00329215686274
net-scale-factor=0.0189601459307
offsets=112.86182266638355;112.86182266638355;112.86182266638355

#onnx-file=/home/jetson-nx/codes/models/facenet/v2_facenet_b16.onnx
model-engine-file=/home/riotu/codes/models/facenet/agx_facenet_dynamic_model.onnx_b16_gpu0_fp16.engine
force-implicit-batch-dim=1
batch-size=16
# 0=FP32 and 1=INT8 2=FP16 mode 
network-mode=2

gie-unique-id=3
operate-on-gie-id=2
operate-on-class-ids=0

#is-classifier=1
#classifier-async-mode=1
network-type=100

output-blob-names= Bottleneck_BatchNorm/batchnorm_1/add_1:0


input-object-min-width=10
input-object-min-height=10
model-color-format=0      

output-tensor-meta=1
#scaling-filter=1
#scaling-compute-hw=0
	
maintain-aspect-ratio=1
#secondary-reinfer-interval=16

#avg_mean =  112.86182266638355  avg_std =  52.742210089081475

Linking Pipline:

streammux.link(pgie)
pgie.link(tracker)
tracker.link(nvvidconv1)
nvvidconv1.link(filter1)
filter1.link(face_recogniser)
face_recogniser.link(nvvidconv)
nvvidconv.link(nvosd)
nvosd.link(tee)
queue1.link(msgconv)
msgconv.link(msgbroker)


face_recogniser_sinkpad = face_recogniser.get_static_pad("src")
if not face_recogniser_sinkpad:
	sys.stderr.write(" Unable to get sink pad of face_recogniser \n")
face_recogniser_sinkpad.add_probe(Gst.PadProbeType.BUFFER, sgie_sink_pad_buffer_probe, 0)

• Hardware Platform: Nano
• DeepStream Version: 5.1
• JetPack Version: 4.5.1

3

We don’t know what happened with your models and codes.

4

Could I send the code to you in private?

5

Can you write an app to reproduce the failure with Nvidia sample model?

6

No, because I am using face recognition classifier and Nvidia does not have one like it

7

Due to law, Nvidia can not provide face recognition function.

If you think it is a deepstream problem which has nothing to do with the model, you can reproduce it with similar model. We have provided several sample classifier models in deepstream samples.

8

What I am sure is that the problem from msgbrocker. Because I test the the model before I move to test4. So I need help for this problem

9

What are these three messages? Is there any person message in these three messages?

What kind of test you have done? With which sample app?

10

Yes

I tested on deepstream_multistream_test app

11

So it is easy to reproduce the crash in your enviroment. Can you debug with test4 and find out which part run into crash? Or do you have other method to tell us how to reproduce the crash with our models?

12

This repo for code and onnx model

13

The model link is not accessible.

14

you have access now

15

Please provide platform information:

**• Hardware Platform (Jetson / GPU)**
**• DeepStream Version**
**• JetPack Version (valid for Jetson only)**
**• TensorRT Version**
**• NVIDIA GPU Driver Version (valid for GPU only)**
16

here

17

I’ve tried your code. It does not work because the scikit-learn package incompatibility. If you think the problem is caused by deepstream, can you provide a clean version which have no other dependency to external packages?

19

I told you the model work fine on deepstream_multistream app. So the problem not becuase of scikit-learn package

20

The code can not run on my platform due to the scikit-learn package incompatibility problem. Can you provide a simple version which does not depends on other 3rd party libraries?

21

on sgie_sink_pad_buffer_probe change this line

face_to_predict_embedding = normalize_vectors(reshape_face_embedding)

to

face_to_predict_embedding = reshape_face_embedding /255
22

Any help @Fiona.Chen ?