LittlePey / SFD

Sparse Fuse Dense: Towards High Quality 3D Detection with Depth Completion (CVPR 2022, Oral)
Apache License 2.0
264 stars 35 forks source link

Inference speed #8

Closed CBY-9527 closed 2 years ago

CBY-9527 commented 2 years ago

The inference speed mentioned in the paper is 10.2 fps. In the codes, the depth complement files have been provided. So how do we get the real inference speed.

LittlePey commented 2 years ago

The latency reported in the paper includes the depth completion network, you can refer to the PENet and TWISE for latency of depth completion network.

CBY-9527 commented 2 years ago

The latency reported in the paper includes the depth completion network, you can refer to the PENet and TWISE for latency of depth completion network.

Thank you for your reply. Does the average processing time per frame of depth estimation network + average time per frame of detection network detection represent the final inference time.

LittlePey commented 2 years ago

Yes