search

uber-research / LaneGCN

[ECCV2020 Oral] Learning Lane Graph Representations for Motion Forecasting
https://arxiv.org/abs/2007.13732
Other
496 stars 131 forks source link

inference time #12

Closed weidezhang closed 3 years ago

weidezhang commented 3 years ago

Sorry I just post a random question here. It appears every object need to go through the network once to predict trajectory. I am wondering about the online inference time of your network (maybe after onnx). Say if you have 20 objects in the map, can that be done in real time (say within couple of mili-seconds)?

kaikaishen commented 3 years ago

I am deploying model to C++ Project.Welcome to exchange.

chenyuntc commented 3 years ago

Hi @weidezhang, that's more like engineering. We didn't try that.

weidezhang commented 3 years ago

thx