How to get classification results which from the classifier model

[422845694]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson NX
• DeepStream Version 6.0
**• JetPack Version (valid for Jetson only)**4.6
• TensorRT Version 8.0

1、The classification model is a customized model, trained with pytorch, and then converted to engine file with trtexec
2、Model input: whole frame (NOT ROI)
3、The classification model inference as primary engine
config text :
[property]
gpu-id=0

preprocessing parameters: These are the same for all classification models generated by TAO Toolkit.

net-scale-factor=0.0078431372549
offsets=0.5;0.5;0.5
model-color-format=0
batch-size=1

labelfile-path=cfgs/ck_cls_label.txt
model-engine-file=cfgs/engine/ck5cls.engine
infer-dims=3;512;512 # where c = number of channels, h = height of the model input, w = width of model input
uff-input-blob-name=input
#uff-input-order=0
output-blob-names=output

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

network-mode=0

process-mode: 2 - inferences on crops from primary detector, 1 - inferences on whole frame

process-mode=1
interval=0
network-type=1 # defines that the model is a classifier.
gie-unique-id=1
classifier-threshold=0.2

[yingliu]

Are you using DeepStream for inference? Thanks to describe the software configs and reproducing steps in details.

[422845694]

1、Use deepstream6 0 read RTSP stream inference
2、pgie is classification model ,which inputs use RTSP full stream(NOT ROI)

i don’t konw how to get the raw output from the classification model

[kesong]

Do you mean you want to “output-tensor-meta” for nvinfer?

Gst-nvinfer — DeepStream 6.0.1 Release documentation (nvidia.com)

[422845694]

Thank you for your reply. This method has been solved. How can I get the complete image of the video stream and convert it to opencv mat format