search

zhouyuchong / gst-nvinfer-custom

Custom gst-nvinfer for alignment in Deepstream
22 stars 12 forks source link

How is complied with c++ #4

Closed NaifOAlharbi closed 1 year ago

NaifOAlharbi commented 1 year ago

Hello, how can I use this custom with deepstream_infer_meta_test.cpp?? Which part should I change in order to apply it?

zhouyuchong commented 1 year ago

@NaifOAlharbi hi, where can i find deepstream_infer_meta_test.cpp.

custom gst-nvinfer is used to replace gst-nvinfer. you can find original gst-nvinfer so file at 

/opt/nvidia/deepstream/deepstream-6.1/lib/gst-plugins/libnvdsgst_infer.so
zhouyuchong commented 1 year ago

@NaifOAlharbi If u want to save the pics, u can run these commands before run urown script.

mkdir -p images/{origin, aligned}

u can modify codes here and this line.

as for the output of retinaface, I'm using the tensorrtx version, which has only one layer output but also contains landmarks. if u want to use urown model, u should modify these codes which extracts info from usermeta-data.

zhouyuchong commented 1 year ago

@NaifOAlharbi yes that's right. I copy the data from gpu and do alignment. then cover data in gpu with aligned result so sgie will use these data to infer.