Different batch size effect the result

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

Please provide complete information as applicable to your setup.

• T4 (Jetson / GPU)
• DeepStream 5.01
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)

Hi all,
I use my mode to run deepstream ,same video and same apllication ,we found the result of detect is different between batch size =1 and batch size =20.

the result in batch size =1 is more than the result in batch size =20.

can you tell me why it happen , I want to use batch size 20, the performance is better than batch size 1.

please tell me how to configure the setting.

thanks

3

I didn’t get here.

4

for example： use batch size 1， the detect box is 20276， but batch size = 20， the number of detect is 20246.

5

I think you can try the model with different BS without deepstream pipeline to check it the issue is caused by the model itself