Simple classifier as Primary GIE

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Is it possible to use a simple binary classifier as the pgie, without a primary detector, or secondary classifiers?

I put together a proof of concept, but the frame meta data has a class_id of -1 and the place where secondary classifiers store attributes is empty.

Thanks!

2

In deepstream 3.0, set is-classifier=1 for pgie config

Deepstream 4.0 will also support this. 4.0 will be released soon, maybe in 3 weeks.

In 4.0, need to set process-mode=1 and network-type=1 for classifiler pgie config.

3

Any ideas why deepstream 3.0 would report a -1 classification from a tensorrt engine?

4

The classification threshold is high. Or the pre-processing parameters are incorrect

5

In gstnvinfer_meta_utils.cpp:

void
attach_metadata_classifier (GstNvInfer * nvinfer, GstMiniObject * tensor_out_object,
    GstNvInferFrame & frame, GstNvInferObjectInfo & object_info)
{
    ...
    object_meta->class_id = -1;
    ...
}

If only a classification model is being used, does class_id ever get set to anything other than -1?

6

Hi mattcarp88,

Please help to open a new topic for this issue. Thanks