FPS goes down in batch processing with default resnet model

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) Jetson Nano
• DeepStream Version 5.1
• JetPack Version (valid for Jetson only) 4.5
• TensorRT Version 7.1
• NVIDIA GPU Driver Version (valid for GPU only)
**• Issue Type( questions, new requirements, bugs)**Questions
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)
I ran the deepstream-app -c /opt/nvidia/deepstream/deepstream-5.1/samples/configs/deepstream-app/source8_1080p_dec_infer-resnet_tracker_tiled_display_fp16_nano.txt
with 4 and 1 input sources with pgie batch size set as 4 & 1 respectively.
I am observing exponential drop of FPS in this case.
This should not happen in batch processing as all 4 frames will be processed in single batch correct or am I wrong?
FPS with 1 source and batch size 1 = **PERF: 63.39 (64.30)
FPS with 4 sources and batch size 4 = **PERF: 19.52 (19.63) 19.52 (19.63) 19.52 (19.63) 19.52 (19.63)

We can achieve 30fps for each source in running default source8_1080p_dec_infer-resnet_tracker_tiled_display_fp16_nano.txt. Are you able to achieve the same performance with the default config file? Would like to make sure we get identical result with the config file.

@DaneLLL Where can I find the original configurations? I have changed some parameters.

I restored with the default configurations, I am getting 30 FPS for multiple inputs but way more for single input.

Do you keep interval=4 in [primary-gie]? GPU in Jetson Nano may not be capable of processing every frame in multi-source use-case.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.