Back-to-back detector with more than one source

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 5.0
• TensorRT Version 7.1.3
• NVIDIA GPU Driver Version (valid for GPU only) CUDA 10.2
Hi Everybody !
I follow back-to-back detector, i run it successfuly. However I want to run with multiple input source. I changed the source code and it work when i run only one source but it dont work when i set more than one source, it raise the below error.


I see that the error in
gst_nvinfer_output_loop function in gstnvinfer.cpp at lines from 1966 to 1981

Do you know that why and how can i fix it ? please help me.

The back-to-back sample can only support one video stream. So the failure is related to your change of multiple source. Can you upload your changed codes or tell us the details of your change?

hi, this is my back_to_back_detector.c that i changed https://drive.google.com/file/d/1uk9ZC3fqQOl5a4LX0d-twzxnmyWgXRHq/view?usp=sharing
I follow multiple souces in deepstream-test3 example.

You created a nvmultistreamtiler element but forgot to add it to your pipeline. It is necessary for multiple sources case. Please check deepstream-test3 codes carefully.

Thanks a lot, it work well.

what is a back-to-back detector?

that is deepstream app example. https://github.com/NVIDIA-AI-IOT/deepstream_reference_apps/tree/master/back-to-back-detectors

does back to back detectors use 2 detectors?

yes, that does.

does it have a python implementation?

i want to run a yolotiny model to detect an object and once again use another yolotiny model for further detections? will this back to back detector be useful?

I dont know but i think it not avaliable.

right, you can build your system and custom it in back-to-back detector example.

1 Like

thank you so much

I downloaded the example file, does the examples file find a car with primary detector and license plate with secondary detector?