compile failed-NvInfer.h cannot found

[MolianWH]

when make, it occurs errors as fallow

I run the command as fallow to clone tensorrt-zoo. git clone --recursive https://github.com/zerollzeng/tensorrt-zoo.git

My TensorRT version is 6.0.1.5. It success when run the TensorRT sample test. But it seems that tiny-tensorrt can not find tensorRT include dir.

Do I need to change TensorRT install path? What's your directory structure?

Thanks!

[MolianWH]

Hi, Are there more detailed installation instructions？I add TensorRT include folders manually， but there are so many errors when compile, such as missing OpenCV headers and missing link libraries etc.

[MolianWH]

Hi, Thanks for providing resources. The problems have been solved. It seems that the installation environment of TensorRT is different.

[zerollzeng]

Sorry for late reply, If you install tensorrt via official pkg file then the so should be in /usr/lib/x86_64-linux-gnu/, and the and the header file was in /usr/include/x86_64-linux-gnu/. if you hava different tensorrt installation path then you should add the path in CMakeLists.txt, I havn't try it before, but it should works. Also if your problem has been solve, can you kindly post your solution so that other people could benifit from you?

[MolianWH]

Here are errors I met and solutions.

1. cmake -D BUILD_PYTHON=ON .. not cmake -D PYTHON_API ON ..
2. error: ‘CV_BGR2RGB’ was not declared in this scope cv::cvtColor(img, rgb, CV_BGR2RGB); solutions: findyolo3test.cpp，and add include file as fallow:

   #include <opencv2/imgproc/types_c.h>

3. errors with TensorRT envs solutions: find CMakeLists.txt in root path, and add include dir and libs dir :

# TensorRT
set(TENSORRT_ROOT "/home/dreamdeck/Downloads/TensorRT-6.0.1.5")
find_path(TENSORRT_INCLUDE_DIR NvInfer.h
  HINTS ${TENSORRT_ROOT} ${CUDA_TOOLKIT_ROOT_DIR}
  PATH_SUFFIXES include)
MESSAGE(STATUS "Found TensorRT headers at ${TENSORRT_INCLUDE_DIR}")
include_directories(${TENSORRT_INCLUDE_DIR})

   link_directories(your_TensorRT_Path)

4. There is also an error when run testopenpose error: terminate called after throwing an instance of 'std::invalid_argument' what(): stoi Aborted (core dumped) solutions: run as

   path/to/testopenpose --prototxt path/to/prototxt --caffemodel path/to/caffemodel/ --save_engine path/to/save_engin --input path/to/input/img --run_mode 1

   run_mode = 1 means setFp16Mode run_mode = 2 means set int8 inference mode there suggests @zerollzeng write to README.md

5. don't generate *.engine file haven't solved.

[zerollzeng]

Thanks for your kindly contribution! I will update docs fo tensorrt-zoo, and if you don't want to generate .engine file, just pass enginefilename with "" empty string

[zerollzeng]

Already update documentation @MolianWH, Please note that tensorrt-zoo is not compatible with latest tiny-tensorrt because updated api

[zerollzeng]

your command is --save instead of --save_engine?

[weilanShi]

Here are errors I met and solutions.

1. cmake -D BUILD_PYTHON=ON ..   not  cmake -D PYTHON_API ON ..

2. error:  ‘CV_BGR2RGB’ was not declared in this scope
   cv::cvtColor(img, rgb, CV_BGR2RGB);
   solutions: find[yolo3test.cpp](./yolov3/yolo3test.cpp)，and add include file as fallow:

#include <opencv2/imgproc/types_c.h>

1. errors with TensorRT envs
   solutions: find [CMakeLists.txt](./CMakeLists.txt) in root path, and add include dir and libs dir :

# TensorRT
set(TENSORRT_ROOT "/home/dreamdeck/Downloads/TensorRT-6.0.1.5")
find_path(TENSORRT_INCLUDE_DIR NvInfer.h
  HINTS ${TENSORRT_ROOT} ${CUDA_TOOLKIT_ROOT_DIR}
  PATH_SUFFIXES include)
MESSAGE(STATUS "Found TensorRT headers at ${TENSORRT_INCLUDE_DIR}")
include_directories(${TENSORRT_INCLUDE_DIR})

   link_directories(your_TensorRT_Path)

1. There is also an error when run testopenpose
   error: terminate called after throwing an instance of 'std::invalid_argument'   what():  stoi Aborted (core dumped)
   solutions:
   run as

path/to/testopenpose --prototxt path/to/prototxt --caffemodel path/to/caffemodel/ --save_engine path/to/save_engin --input path/to/input/img --run_mode 1

run_mode = 1 means setFp16Mode run_mode = 2 means set int8 inference mode there suggests @zerollzeng write to README.md

1. don't generate *.engine file
   haven't solved.

I met the same error : terminate called after throwing an instance of 'std::invalid_argument' what(): stoi Aborted

and there is my command: ./testopenpose --prototxt pose_deploy.prototxt --caffemodel pose_iter_584000.caffemodel --save_engine ./save_engine --input test.jpg could you tell me how to solve it?

[zerollzeng]

I met the same error : terminate called after throwing an instance of 'std::invalid_argument' what(): stoi Aborted

and there is my command: ./testopenpose --prototxt pose_deploy.prototxt --caffemodel pose_iter_584000.caffemodel --save_engine ./save_engine --input test.jpg could you tell me how to solve it?

https://github.com/zerollzeng/tensorrt-zoo/blob/238c83103bb97aa799ebc67f4dee583e1939fe20/openpose/testopenpose.cpp#L48

arguments are missing, please take a look at testopenpose.cpp