search

zhangjinsong3 / plate_metric.libtorch

A sample for libtorch c++ API,use opencv Mat or GpuMat as input.
8 stars 0 forks source link

Question about Yolov3-tenosrrt project. #1

Open MuhammadAsadJaved opened 3 years ago

MuhammadAsadJaved commented 3 years ago

Hi, I have a question about your Yolov3-TensorRT project. is it possible to talk over wechat? My WeChat is : 13621297680

zhangjinsong3 commented 3 years ago

Hi, I have a question about your Yolov3-TensorRT project. is it possible to talk over wechat? My WeChat is : 13621297680

That repo is too old. You can reference tensorRT samples, which provides an yolov3->onnx->trt sample that do not need plugin!

MuhammadAsadJaved commented 3 years ago

I have a different question. Seems like you are expert in TensorRT. I have a Yolov3 that basically takes two inputs (visual image and infrared image). then perform features extraction, features fusion and finally object detection. it's almost the same as Yolov3, just changed for two inputs. I have converted this model .pb > .onnx > .trt. Now I just want to ask how I can use htis .trt model for the interface? How I can change TensorRT/samples/python/yolov3_onnx for this purpose?

On Wed, Oct 28, 2020 at 1:27 PM zhangjinsong3 notifications@github.com wrote:

Hi, I have a question about your Yolov3-TensorRT project. is it possible to talk over wechat? My WeChat is : 13621297680

That repo is too old. You can reference tensorRT samples, which provides an yolov3->onnx->trt sample that do not need plugin!

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/zhangjinsong3/plate_metric.libtorch/issues/1#issuecomment-717707249, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG4GR5AWG4GF7QIMUPHJUN3SM6TT3ANCNFSM4TB2LPSA .

zhangjinsong3 commented 3 years ago

I have a different question. Seems like you are expert in TensorRT. I have a Yolov3 that basically takes two inputs (visual image and infrared image). then perform features extraction, features fusion and finally object detection. it's almost the same as Yolov3, just changed for two inputs. I have converted this model .pb > .onnx > .trt. Now I just want to ask how I can use htis .trt model for the interface? How I can change TensorRT/samples/python/yolov3_onnx for this purpose? … On Wed, Oct 28, 2020 at 1:27 PM zhangjinsong3 @.***> wrote: Hi, I have a question about your Yolov3-TensorRT project. is it possible to talk over wechat? My WeChat is : 13621297680 That repo is too old. You can reference tensorRT samples, which provides an yolov3->onnx->trt sample that do not need plugin! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AG4GR5AWG4GF7QIMUPHJUN3SM6TT3ANCNFSM4TB2LPSA .

Seems that your question is on 'How to take 2 input2 into tensorRT engine'. I haven`t meet this problem before. In the yolo-tensorrt repo, there is only 1 input and have 3 outputs. In my opinion, if engine do not support 2 inputs (Although it probably supports this), u can split your model. Let it become 2 sub models extracting feature of the 2 inputs and the concat them, input the concated features to 3th model.