Error in Deepstream Objectdetector-SSD while integrating Custom SSD model

Accelerated Computing Intelligent Video Analytics DeepStream SDK
,
#1

• Hardware Platform - Jetson
• DeepStream Version - 5.0
• JetPack Version - 4.4.1
• TensorRT Version -7.1.3

Hi Team,

I am facing an error when I was trying to use my own cusom SSD model in the deepstream objectdetector-SSD sample.

its a ONNX model converted from pytorch checkpoint.

I am attaching the error log here. The error was related to NMS layer.

error_with_my_ssd_model
error_with_my_ssd_model1214×852 217 KB

I just gone through the network arcchitecture and cust_bbox_parser.cpp. There is not NMS layer in my ssd model.

my_network_output_arch
my_network_output_arch1188×829 67.9 KB

How can i use my custom SSD model with the Objectdetector-SSD script.
I am attaching my config files too.
config_infer_primary_ssd.txt (3.9 KB) deepstream_app_config_ssd.txt (2.5 KB)

@Fiona.Chen en , @kayccc @WayneWWW can anybody help me in this?

#5

since your output layer is different from NMS, you need to customize your own parser for your model.

Now, for Objectdetector-SSD, its output parser is in objectDetector_SSD/nvdsinfer_custom_impl_ssd/nvdsparsebbox_ssd.cpp