Integrate YOLOv5 as pgie and OCRNet as sgie

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Please provide the following information when requesting support.

• Hardware : 4070ti

Hi,
I’m trying to integrate YOLOv5 as pgie and OCRNet as sgie. How to generate the custom-lib-path for OCRNet? I have checked NVIDIA-Optical-Character-Detection-and-Recognition-Solution. Here OCDNEt is used for detection. I want to use YOLOV5. Any help is highly appreciated. @Morganh

2

It is needed to do implementation in the github NVIDIA-Optical-Character-Detection-and-Recognition-Solution, to make it work by using YOLOv5 engine and OCRnet engine.

Then, you can go to use deepstream as well, by default, the available app is in deepstream_tao_apps/apps/tao_others/deepstream-nvocdr-app at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub.

3

Hi, thanks for your response. I wan t to know whether the other way i.e. using the .so file for OCRNet and integrate it as sgie as shown in the deepstream sample app. How to generate the .so file for OCRNet?

4

Currently there is not standalone lib for OCRNet. For the github. GitHub - NVIDIA-AI-IOT/NVIDIA-Optical-Character-Detection-and-Recognition-Solution: This repository provides optical character detection and recognition solution optimized on Nvidia devices.. After compiling, there is libnvocdr.so.

1 Like