ONNX classification on deepstream test 5

I’ve trained ONNX model, it’s working fine as object detection with yolo v2, but not as classification.

How can I solve this issue, and include the classification in azure message broker?

Also how can I integrate deepstream test 5 with openALPR?

Hi,

Would you mind to share more information about your issue?
Is your onnx model a detector or a classifier?

To use message broker, please check this document for more detail:
https://docs.nvidia.com/metropolis/deepstream/plugin-manual/index.html#page/DeepStream_Plugin_Manual%2Fdeepstream_plugin_details.02.15.html

For openALPR, you can start from this GitHub:

Thanks.

I’ve tried both ONNX models detector or a classifier, and both are not working as classifiers.

Regarding ALPR, where can I find nvgstiva-app, and if there’s any sample how to intgrate it with deepstream-test5

Hi,

May I know how do you integrate your onnx classifier to test5?
And would you mind share the detail of your error log?

Thanks.

I didn’t change anything, I just added below to the configs, and it’s working fine for object detection, but when I tried to use Classification, there’s no error, but there’s no output. Also please let me know if there’s any sample how to integrate with ALPR as I can’t find nvgstiva-app!!

[secondary-gie2]
enable=1
gpu-id=0
gie-unique-id=7
operate-on-gie-id=1
operate-on-class-ids=0
batch-size=30
config-file=config_Classification.txt
labelfile-path=…/custom_models/Classification/labels.txt
onnx-engine-file=…/custom_models/Classification/model.onnx_b30_fp32.engine
nvbuf-memory-type=0

[property]
gpu-id=0
net-scale-factor=1
#0=RGB, 1=BGR
model-color-format=1
process-mode=1
onnx-file=…/custom_models/Classification/model.onnx
labelfile-path=…/custom_models/Classification/labels.txt

0=FP32, 1=INT8, 2=FP16 mode

network-mode=0
num-detected-classes=9
gie-unique-id=7
is-classifier=0
operate-on-gie-id=1
operate-on-class-ids=0
maintain-aspect-ratio=1
#output-blob-names=output-blob-names=coverage;bbox
parse-bbox-func-name=NvDsInferParseCustomYoloV2
custom-lib-path=…/custom_models/Classification/libnvdsinfer_custom_impl_Yolo_Custom_Vision.so

Hi,

The OpenALPR integration is based on previous Deepstream SDK.
You will need to rewrite the sample for Deepstream 4.0.

To give further suggestion, would you mind to share what kind of error do you meet with onnx classifier?
Please noticed that you will need to update following property for a classifier:

# 0: Detector, 1: Classifier,2: Segmentation
network-type=0
is-classifier=1

More detail can be found in our document here:
https://docs.nvidia.com/metropolis/deepstream/plugin-manual/index.html#page/DeepStream_Plugin_Manual%2Fdeepstream_plugin_details.02.01.html%23wwpID0E0QX0HA

Thanks.