How to create custom logic in msgConv and broker?

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Hi,

i am using yolov3-tiny to track the person who doesn’t ware safety cap.

when i get the infer info (w,h,x,y), how can i crop the img? i mean i just want to save the img[x:x+w,y:y+h] to my nano desktop. and i want to resend the buffer to the rtmp stream. such as (rtmp://localhost/live)

i try to code it in your msgconv and broker file. but there is little document.

that’s too hard , please help!!!

if someone know any helpful resources,please tell me , thank u :))))

2

Here is cuda kernel of image crop: https://devtalk.nvidia.com/default/topic/1032756/jetson-tx1/what-is-the-easiest-way-to-crop-an-image-on-the-gpu-/post/5254287/#5254287

Opencv also has crop function.

gstnvinfer.cpp:line1192 uses NvBufSurfTransform()

You can replace your own rtmp protocol in msgbroker

3

Hi,thanks for your quickly reply.
where should i apple this useful .cu fucntion?
in your gst-nvmsgconv? or gst-nvmsgbroker ?

i’m using deepstream4.0 on jetson nano

4

Could anyone help me?
I wanna using aio.h to save the crop image