Cannot obtain the pgie classifier result using nvpreprocess

Please provide complete information as applicable to your setup.

**• Hardware Platform (Jetson / GPU)GPU
**• DeepStream Version 6.0

If the input of a pgie classifier is the input tensor meta data(ROI) processed by a nvpreprocess, the classification result will not obtained from the NvDsObjectMeta with class_id= -1. How could I obtain the result?

I think the code for that is very strange because if I set the process mode of this classifier is 1(on full frame), everything works very well.

What is your expectation?

I would like to obtain the classification result from the NvDsClassifierMeta in the NvDsObjectMeta, even I use a nvpreprocess to preprocess the input data for nvinfer. Now I can only use the global variable to obtain these result from my custom library for postprocess, it is quite strange for my project.

But you said it works fine in full frame mode. Is it the expect result?

That is just a test.I would like to use it after nvpreprocess because it will make the model performance much better.

Is it possible to reproduce it with DS sample?

Yep, you can use the deepstream-app, just setting a pgie classifier config and a preprocess config, then turn on input-tensor-meta=1 for the primary-gie group in your config of deepstream-app, you will find that you cannot print the object class id=-1 from the all _bbox_generated function in deepstream-app-main.c. However, if you turn off it, you can see the normal result.

i meet the same error, any solution?

Sorry to later response. Is it one issue for you?

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.

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