Unable using either docker file to build my own model

[jiver54]

I am trying to train my own caffee model. I have a pre existing set of images and annotations which were created using labelImg.
I installed the docker file:
sudo docker run -ti --name=simcamsdk -v /home/myname/:/home/myname/ simcam/simcamsdk:v1.0 bash

I copied the images to to:
/SIMCAM_SDK_v1.0.0/train/data/Images_xmls/JPEGImages
and the XML annotations to:
/SIMCAM_SDK_v1.0.0/train/data/Images_xmls/Annotations
Executed: python create_txt.py
This created the appropriate text files
I modified the
/SIMCAM_SDK_v1.0.0/train/data/lmdb_files/labelmap.prototxt
file as such:
item {
name: "WeedyLeft"
label: 0
display_name: "WeedyLeft"
}
item {
name: "WeedyRight"
label: 1
display_name: "WeedyRight"
}
item {
name: "FlankLeft"
label: 2
display_name: "FlankLeft"
}
item {
name: "FlankRight"
label: 3
display_name: "FlankRight"
}

I ran the script: ./create_list.sh
This threw the following error:.....
./create_list.sh: line 17: [: missing ]' Create list for Images_xmls trainval... ./create_list.sh: line 17: [: missing ]'
Create list for Images_xmls test...
WARNING: Logging before InitGoogleLogging() is written to STDERR
F0923 08:10:51.719509 63 get_image_size.cpp:110] This tool requires OpenCV; compile with USE_OPENCV.
*** Check failure stack trace: ***
./create_list.sh: line 8: 63 Aborted (core dumped) $caffe_root/build/tools/get_image_size $root_dir $dst_file $bash_dir/$dataset"_name_size.txt"

Then I tried
sudo docker run --runtime=nvidia -ti --name=simcam -v /home/your_username:/home/your_username simcam/caffe-ssd:gpu bash
This would only run with the --runtime=nvidia removed..
The SDK is not in this docker file.. So I started the install
I had to
get-apt install git
git clone https://github.com/simshineaicamera/SIMCAM_SDK.git
cd SIMCAM_SDK
cd SIMCAM_SDK/tools/arm_toolchain
chmod +x cross.v300.install
./cross.v300.install
apt-get install lib32z1
apt-get install lib32stdc++6-4.8-dbg
./install-ncsdk.sh
vi ~/.bashrc
(Added this line to the bottom of the file)
export PATH="/opt/hisi-linux/x86-arm/arm-hisiv300-linux/target/bin:$PATH"
source ~/.bashrc
arm-hisiv300-linux-gcc -v
cd $SIMCAM_SDK/tools/mv_toolchain/model_conversion/
./install-ncsdk.sh

I got the following error...
Movidius Neural Compute Toolkit Setup.
Checking OS and version...
Your current combination of linux distribution and distribution version is not officially supported!

So I have failed with BOTH docker files... sigh... any ideas?

[jiver54]

I have fixed one of the problems when I run the ./create_list.sh: .... I space is needed on line 17 i.e.
instead of if [ $dataset == "test"] ... it should be if [ $dataset == "test" ] ... with a space.

The error now is ..
Create list for Images_xmls trainval...
WARNING: Logging before InitGoogleLogging() is written to STDERR
F0924 01:13:32.694209 210 get_image_size.cpp:110] This tool requires OpenCV; compile with USE_OPENCV.
*** Check failure stack trace: ***
./create_listFIX.sh: line 8: 210 Aborted (core dumped) $caffe_root/build/tools/get_image_size $root_dir $dst_file $bash_dir/$dataset"_name_size.txt"

I Installed version3 of openCV
I used the instructions on
https://gist.github.com/arundasan91/b432cb011d1c45b65222d0fac5f9232c
This failed except when I did the following:

cd /usr/lib/x86_64-linux-gnu
sudo ln -s libhdf5_serial.so.8.0.2 libhdf5.so
sudo ln -s libhdf5_serial_hl.so.8.0.2 libhdf5_hl.so
and as well as the Makefile.config flags I also set

INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial/ and
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial/
It then compiled..
That is now working properly, except now I get.
./create_list.sh: line 46: /opt/movidius/ssd-caffe/build/tools/get_image_size: No such file or directory
I will keep trying...

[jiver54]

Reloaded the whole thing again to make sure amongst the debugging I didn't break something..
So the non-GPU docker file.. if you are to use that there is a syntax error in the "create_list.sh" file and caffe has been compiled without the openCV switch being set. This switch needs to be set so "create_list.sh" can run. They are definite errors. (OpenCV is not actually installed) Also when you have a Docker session labelImg does not run. It throws an error "labelImg: cannot connect to X server localhost:1.0"...
"