search

TRI-ML / dd3d

Official PyTorch implementation of DD3D: Is Pseudo-Lidar needed for Monocular 3D Object detection? (ICCV 2021), Dennis Park*, Rares Ambrus*, Vitor Guizilini, Jie Li, and Adrien Gaidon.
MIT License
451 stars 74 forks source link

Export ONNX model #11

Open williamhyin opened 2 years ago

williamhyin commented 2 years ago

Hello

Could you share your onnx export script, which can help us a lot.

Thanks!

dennis-park-TRI commented 2 years ago

Hello @williamhyin good point. Currently, I don't have a bandwidth for doing this, but I will keep this in my backlog.

williamhyin commented 2 years ago

Hello @williamhyin good point. Currently, I don't have a bandwidth for doing this, but I will keep this in my backlog.

Hi ,

Thanks for your reply. II noticed that DD3D uses the detectron2 architecture. Detectron2-onnx conversation is based on Caffe2Tracel. Is it possible to use the deployment API of detectron2 to perform ONNX conversion?

link: https://detectron2.readthedocs.io/en/latest/_modules/detectron2/export/api.html#Caffe2Model tutorial: https://detectron2.readthedocs.io/en/latest/tutorials/deployment.html

But I faced the problem when i use onnx-export command: 

onnxModel = export_onnx_model(cfg, model, inputs)
onnx.save(onnxModel, "/content/deploy.onnx")

bug: assert isinstance(cfg, CfgNode), cfg

It seems like the dd3d-cfg is not the official format cfg file. And the inputs from dataloder is more complex. Could you give me some idea or suggestion?

Thanks! I look forward to hearing from you.

JensenGao commented 2 years ago

Hello @williamhyin good point. Currently, I don't have a bandwidth for doing this, but I will keep this in my backlog.

Hi ,

Thanks for your reply. II noticed that DD3D uses the detectron2 architecture. Detectron2-onnx conversation is based on Caffe2Tracel. Is it possible to use the deployment API of detectron2 to perform ONNX conversion?

link: https://detectron2.readthedocs.io/en/latest/_modules/detectron2/export/api.html#Caffe2Model tutorial: https://detectron2.readthedocs.io/en/latest/tutorials/deployment.html

But I faced the problem when i use onnx-export command:

onnxModel = export_onnx_model(cfg, model, inputs)
onnx.save(onnxModel, "/content/deploy.onnx")

bug: assert isinstance(cfg, CfgNode), cfg

It seems like the dd3d-cfg is not the official format cfg file. And the inputs from dataloder is more complex. Could you give me some idea or suggestion?

Thanks! I look forward to hearing from you.

Have you saved you problem?