-
Notifications
You must be signed in to change notification settings - Fork 51
Open
Description
Hi @CarkusL ,
I came across your repo and I am trying out something similar. It helped me a lot in understanding things. I am trying to serve the PointPillar model to an inference engine with ONNX runtime on it. For that I need the ONNX model. But I noticed that the Pytorch model in your repo includes the bbox_head and predicts the boxes directly which are then filtered with NMS. However, the ONNX model does not have this bbox_head included. Did I understand this correctly?
Also the inputs to the ONNX model are: input.1 with shape [1, 10, 30000, 20] and indices_input with shape [1, 30000, 2]. How do I feed the input from the voxel generator to such a model? Any help would be appreciated.
Metadata
Metadata
Assignees
Labels
No labels