search

Sense-GVT / Fast-BEV

Fast-BEV: A Fast and Strong Bird’s-Eye View Perception Baseline
Other
568 stars 85 forks source link

更新deploy代码 #40

Open yhl41001 opened 1 year ago

yhl41001 commented 1 year ago

您好，会更新deploy代码吗？

cyn-liu commented 1 year ago

你好，请问你完成部署了吗？我现在也在解决这个问题，但不知道如何用tensorrt加速这个模型，因为模型中含有Project2Dto3D这个层，tensorrt识别不了，如何利用作者提供的.cu转成tensorrt可识别的插件？

Ignite616 commented 11 months ago

你好，请问你完成部署了吗？我现在也在解决这个问题，但不知道如何用tensorrt加速这个模型，因为模型中含有Project2Dto3D这个层，tensorrt识别不了，如何利用作者提供的.cu转成tensorrt可识别的插件？

Hello, did you achieve the speed in the paper before deployment? I use the M1 model in the paper, and the inference speed on the GTX 3090 is 1 frame per second.

cyn-liu commented 11 months ago

你好，请问你完成部署了吗？我现在也在解决这个问题，但不知道如何用tensorrt加速这个模型，因为模型中含有Project2Dto3D这个层，tensorrt识别不了，如何利用作者提供的.cu转成tensorrt可识别的插件？

Hello, did you achieve the speed in the paper before deployment? I use the M1 model in the paper, and the inference speed on the GTX 3090 is 1 frame per second.

I didn't finish Fast-BEV deployment, and I don't need do this task now. but has someone already finish it. you can refer to the link. https://github.com/cyn-liu/FastBEV https://github.com/cyn-liu/FastBEV-TensorRT