ONNX support

[Saksham Sinha]

This repository is a great work. Thank you for your work.

I was wondering if you can add ONNX support to your model. Currently, your model gives following error while exporting to ONNX. I am using resnet50-SSD with VOC dataset format.

Change made to demo.py at line 148 are-
dummy_input = torch.randn(1, 3, 300, 300)
output_names = ["outputs"]
torch.onnx.export(net, dummy_input, "resnet50ssd.onnx", verbose=True, output_names=output_names)

Command used to run demo.py-
python demo.py --cfg ./configs/ssd_res50_voc.yaml --weights ./weights/ssd_res50_epoch_50_300.pth --images ./images --save_folder ./output

Error-
Traceback (most recent call last):
File "demo.py", line 174, in
main()
File "demo.py", line 140, in main
torch.onnx.export(net, dummy_input, "resnet50ssd.onnx", verbose=True, output_names=output_names)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/onnx/init.py", line 132, in export
strip_doc_string, dynamic_axes)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/onnx/utils.py", line 64, in export
example_outputs=example_outputs, strip_doc_string=strip_doc_string, dynamic_axes=dynamic_axes)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/onnx/utils.py", line 329, in _export
_retain_param_name, do_constant_folding)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/onnx/utils.py", line 213, in _model_to_graph
graph, torch_out = _trace_and_get_graph_from_model(model, args, training)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/onnx/utils.py", line 171, in _trace_and_get_graph_from_model
trace, torch_out = torch.jit.get_trace_graph(model, args, _force_outplace=True)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/jit/init.py", line 256, in get_trace_graph
return LegacyTracedModule(f, _force_outplace, return_inputs)(*args, **kwargs)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/nn/modules/module.py", line 547, in call
result = self.forward(*input, **kwargs)
File "/home/testuser/venvs/refpy368shared/lib/python3.6/site-packages/torch/jit/init.py", line 325, in forward
torch._C._tracer_exit(tuple(out_vars))
RuntimeError: output of traced region did not have observable data dependence with trace inputs; this probably indicates your program cannot be understood by the tracer.