Problems when install Darknet for jetson-nano

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
1

Hi , now I use this image with my jetson-nano :
[url]https://developer.nvidia.com/embedded/dlc/jetson-nano-dev-kit-sd-card-image[/url]

And then , I had already insert command below in " ~/.bashrc " :

================== in ~/.bashrc ==================================

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

===================================================================

I try to install darknet ( follow the step to install : Darknet: Open Source Neural Networks in C ), but has some errors bellow:

===================== error ========================================

g++: error: obj/convolutional_kernels.o: No such file or directory
g++: error: obj/activation_kernels.o: No such file or directory
g++: error: obj/im2col_kernels.o: No such file or directory
g++: error: obj/col2im_kernels.o: No such file or directory
g++: error: obj/blas_kernels.o: No such file or directory
g++: error: obj/crop_layer_kernels.o: No such file or directory
g++: error: obj/dropout_layer_kernels.o: No such file or directory
g++: error: obj/maxpool_layer_kernels.o: No such file or directory
g++: error: obj/network_kernels.o: No such file or directory
g++: error: obj/avgpool_layer_kernels.o: No such file or directory
Makefile:143: recipe for target ‘darknet’ failed
make: *** [darknet] Error 1

=====================================================================

Can tell me how to solve theese error?

Thanks a lot !

2

Hi,

After updating SD card image, have you flashed and installed the package with sdkmanager?
[url]https://developer.nvidia.com/embedded/jetpack[/url]

Thanks.

3

I use your website to download sdkmanager , and when I tried to install , there had a problem

=========== problem ========================

dpkg: error processing archive sdkmanager_0.9.12-4180_amd64.deb (–install):
package architecture (amd64) does not match system (arm64)
Errors were encountered while processing:
sdkmanager_0.9.12-4180_amd64.deb

============================================

my platform is jetson-nano developer kit .

How can I do next step ?

Thanks.

4

Hi,

Please noticed that sdkmanager is executed on the host device.
Thanks.

5

Sorry for that I don’t notice.

I have done that use another host to flash jetson-nano developer kit.

but still have the same problem , and one is different

===================== error ========================================

gcc: error: obj/convolutional_kernels.o: No such file or directory
gcc: error: obj/deconvolutional_kernels.o: No such file or directory
gcc: error: obj/activation_kernels.o: No such file or directory
gcc: error: obj/im2col_kernels.o: No such file or directory
gcc: error: obj/col2im_kernels.o: No such file or directory
gcc: error: obj/blas_kernels.o: No such file or directory
gcc: error: obj/crop_layer_kernels.o: No such file or directory
gcc: error: obj/dropout_layer_kernels.o: No such file or directory
gcc: error: obj/maxpool_layer_kernels.o: No such file or directory
gcc: error: obj/avgpool_layer_kernels.o: No such file or directory
Makefile:82: recipe for target ‘libdarknet.so’ failed
make: *** [libdarknet.so] Error 1

=====================================================================

The last error become " Makefile:82: recipe for target ‘libdarknet.so’ failed " .

And then ,there commands below are already in " ~/.bashrc "

============== below in ~/. bashrc ==================================

export CUBA_HOME=/usr/local/cuda
export PATH=${PATH}:/usr/local/cuda/bin
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/local/cuda/lib64
source /opt/ros/melodic/setup.bash

====================================================================

And , I can’t find any files like " convolutional_kernels.o " , " deconvolutional_kernels.o " … in

" usr/local/cuda-10.0/include " . Looks likes unusual.

Thanks for answer patiently .

6

Hi,

I can build darnet without issue.
Here is my steps for your reference:

1. Flash Nano and install packages with sdkmanager.

2. Build darnet

$ git clone https://github.com/pjreddie/darknet.git
$ cd darknet
$ make

Thanks.

7

I have done it too! And I think I found the problem .

I edit the parameter of Makefile in darknet , theese parameter can make , see below

============= parameter of Makefile in darknet can make ====================

GPU=1
CUDNN=1
OPENCV=0
OPENMP=0
DEBUG=0

====================================================================

And then , If I use the parameter below , would show the errors like " gcc: error: obj/convolutional_kernels.o: No such file or directory " …

============== parameter of Makefile in darknet can not make =======

GPU=1
CUDNN=1
OPENCV=1
OPENMP=0
DEBUG=0

====================================================================

So I have two question want to know :

  1. It means that opencv ( version = 3.3.1 )in jetson-nano , can not support Deep Neural Networks module?

  2. Could I use darknet in jetson-nano with GPU , if I use the parameter " GPU=1 " and " CUDNN=1 " to make?

8

Hi,

1. We can build darknet with default openCV 3.3.1 without issues.
Here is our Makefile option for your reference:

diff --git a/Makefile b/Makefile
index 63e15e6..5900d0a 100644
--- a/Makefile
+++ b/Makefile
@@ -1,6 +1,6 @@
-GPU=0
-CUDNN=0
-OPENCV=0
+GPU=1
+CUDNN=1
+OPENCV=1
 OPENMP=0
 DEBUG=0

2. Default is already GPU mode.

Thanks.

9

I’ll try again these steps , amd make sure these steps I use are all correct.

( I think maybe I have the wrong step )

Thanks for your answer ~ ! !

10

I have already done it. Thanks for help!