search

Sense-GVT / Fast-BEV

Fast-BEV: A Fast and Strong Bird’s-Eye View Perception Baseline
Other
596 stars 91 forks source link

How to test speed on my own device? #32

Open exiawsh opened 1 year ago

exiawsh commented 1 year ago

Hi, thanks for the great work ! I want to test speed on my own RTX3090 with fp32. Which file do I need to run and where do I modify the config? We want to cite your method and make a fair comparison.

exiawsh commented 1 year ago

We would appreciate it if you could provide the speed on RTX3090 (fp32 pytorch version).

exiawsh commented 1 year ago

I have tested the speed of your resnet-34 model with a single frame on RTX3090. I only got 20 FPS, is that normal?

JudasDie commented 1 year ago

I have tested the speed of your resnet-34 model with a single frame on RTX3090. I only got 20 FPS, is that normal?

I have tested their M0 model (Res18), only 3.3FPS （for the temporal version with 4 frame).

exiawsh commented 1 year ago

I have tested the speed of your resnet-34 model with a single frame on RTX3090. I only got 20 FPS, is that normal?

I have tested their M0 model (Res18), only 3.3FPS （for the temporal version with 4 frame).

Thank you for the result. I set the number of frame to 1, And get the above results. It seems that the pytorch version of the code is not efficient. The speed reported in the paper is mainly due to onnx/tensorrt…

JudasDie commented 1 year ago

I have tested the speed of your resnet-34 model with a single frame on RTX3090. I only got 20 FPS, is that normal?

I have tested their M0 model (Res18), only 3.3FPS （for the temporal version with 4 frame).

Thank you for the result. I set the number of frame to 1, And get the above results. It seems that the pytorch version of the code is not efficient. The speed reported in the paper is mainly due to onnx/tensorrt…

Have you already transferred their model to TRT?

exiawsh commented 1 year ago

I have tested the speed of your resnet-34 model with a single frame on RTX3090. I only got 20 FPS, is that normal?

I have tested their M0 model (Res18), only 3.3FPS （for the temporal version with 4 frame).

Thank you for the result. I set the number of frame to 1, And get the above results. It seems that the pytorch version of the code is not efficient. The speed reported in the paper is mainly due to onnx/tensorrt…

Have you already transferred their model Have not yet, I don't have this plan now.