What is the purpose of squeezenet_ssd_lite.py found in "jetson-inference/python/training/detection/ssd/vision/ssd/"

And the same question for the rest of the python files with architecture names.

Thanks in advance.


You can find the corresponding document in the below link:


It is an alternate architecture that came from the original upstream author. I have only tested SSD-Mobilenet-v1 through that project and found that it works well with exporting to ONNX and importing to TensorRT.

Have you tested the other architectures in your code for exporting to ONNX?

At the time, SSD-Mobilenet-v1 was the only one working with PyTorch ONNX export + ONNX import into TensorRT.