search

cv-core / tensorrt_ros

12 stars 7 forks source link

Performance and accuracy #1

Closed chentairan closed 4 years ago

chentairan commented 4 years ago

Hi All,

As mentioned in your paper, you use Nvidia xavier to run TensorRT for network inference. I would like to ask if you use int8 and what is the accuracy of int8's YoloV3. And have you used two dla to speed up inference ?

Best, Tairan Chen

Sibozhu commented 4 years ago

Hi Tairan, Thank you for your interest in our project.

if you use int8

Yes, we do use int8.

what is the accuracy of int8's YoloV3

We never actually measured it, but it performs similarly to our YoloV3 without quantization.

have you used two dla to speed up inference

Yes

chentairan commented 4 years ago

Thank you for your reply !

Best, Tairan Chen