Threshold used in lprnet inference sgie file does not work

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) jetson
• DeepStream Version 5.1
• JetPack Version (valid for Jetson only) 4.5.1
• TensorRT Version 7
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs) lprnet inference issue
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)

I have retrained lprnet and using for ocr. It is working , but the threshold is not working, I am unable to make the detection stricter using threshold. Please find my sgie file,
property]
gie-unique-id=2
network-type=1 # classifier
process-mode=2 # secondary
operate-on-gie-id=1
operate-on-class-ids=0

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

network-mode=2
batch-size=4
tlt-encoded-model=us_lprnet_baseline18_deployable.etlt
model-engine-file=lpr_us_onnx_b4.engine
labelfile-path=labels.txt
tlt-model-key=nvidia_tlt
custom-lib-path=libnvinfer_custom_lpr_parser.so
parse-classifier-func-name=NvDsInferParseCustomNVPlate

num-detected-classes=3
output-blob-names=output_bbox/BiasAdd;output_cov/Sigmoid
net-scale-factor=0.00392156862745098
#0=RGB 1=BGR 2=GRAY
model-color-format=0
gpu-id=0
interval=0

#workspace-size=1024

[class-attrs-all]
threshold=.99
#classifier-threshold=0.99
#pre-cluster-threshold=0.99
#post-cluster-threshold=0.99

Tried all these different threshold, but none works, please help.

3

if parse-classifier-func-name is set, you need to use classifier-threshold in the custom NvDsInferParseCustomNVPlate, which has a parameter float classifierThreshold.

4

Hi, Thanks for your reply. So do I need to set a value for classifierThreshold at nvinfer_custom_lpr_parser.cpp file?
Cant I set it from the sgie config file?
However I already tried setting classifier-threshold in sgie config file and that did not make any impact or difference, so understood it did not work.
Can you advice.

5

No, becuase you are using custom NvDsInferParseCustomNVPlate. you need to use classifier-threshold in the custom NvDsInferParseCustomNVPlate, which has a parameter float classifierThreshold.

6

Hi,
Thanks a lot.
Can you suggest, How I can use the standard parse classifier and control it from sgie file?

7

please find parse-classifier-func-name in nvinfer. If not set, Gst-nvinfer uses the internal parsing function for softmax layers.

8

Hi,

Thanks a lot. I am still struggling, I am attaching the custom parser file, I tried to understand how to apply the filer in the custom parser cpp file, I could not find where classifier_threshold is used. Can you please have a look at the cpp find and suggest.
nvinfer_custom_lpr_parser.cpp (5.0 KB)

9

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

you can use the parameter “float classifierThreshold” in function NvDsInferParseCustomNVPlate. it should be same with classifier-threshold.

10

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