python yolov3_to_onnx.py for error. about trt_yolov3 HOT 3 CLOSED

to by pass this
Bad node spec: input: "085_convolutional_lrelu" output: "086_upsample" name: "086_upsample" op_type: "Upsample" attribute { name: "mode" s: "nearest" type: STRING } attribute { name: "scales" floats: 1 floats: 1 floats: 2 floats: 2 type: FLOATS }
i have mentioned the ONNX version is different in yolov3 to onnx and onnx to TensorRT engine.

try pip install wget onnx==1.3.0 or 1.2.3

from trt_yolov3.

Since this is the OLD example provided by Nvidia, the code based on very old library that they never mentioned.
Refer to forum, the dependency of TensorRT , python2 , onnx , CUDA have specified version..

Please follow README and pull the container...
If you wanna install tensorRT 5 without container , here's the link

The following env has been proved (for yolov3 to ONNX):

I tried convert onnx to TensorRT 6.1.5 with 19.12-py3 engine and works so well.

Otherwise, i found some models have bugs in TensorRT 7 that cause the wrong inference result.. so i dont recommend TRT7 in this moment.

from trt_yolov3.

Hi, It's working

from trt_yolov3.