I'm reading the Intel® Movidius™ Neural Compute Stick Blog Learn how to design and develop intelligent edge devices with the Neural Compute Stick (NCS). (https://movidius.github.io/blog/deploying-custom-caffe-models/）
Now I want to re-build and try the trial step by step.
First, download the source code and helper scripts from NCAppZoo.
mkdir -p ~/workspace
git clone https://github.com/ashwinvijayakumar/ncappzoo
git checkout dogsvscats
But I couldn't find the diretion: dogsvscats
Is there anybody could help me about this ? Thanks!
Hi @ideallyworld ,
I think you need to be in the ncappzoo cloned repository when executing the
git checkout dogsvscatscommand, not sure if that is what you are referring to when you asked "But I couldn't find the diretion: dogsvscats". I've tried this and it successfully switched to the branch "dogsvscats", I can see also the
<<path-to>path-to>/ncappzoo/apps/dogsvscatsdirectory. Hope this helps.
Hi @Luis_at_Intel ，
Thank you very much for your help. As you said it need be in the ncappzoo cloned repository when executing the git checkout dogsvscats command . And I also see the <path-to>/ncappzoo/apps/dogsvscats directory just as follow:
root@ubuntu:~/workspace# git clone https://github.com/ashwinvijayakumar/ncappzoo
Cloning into 'ncappzoo'…
remote: Enumerating objects: 2980, done.
remote: Total 2980 (delta 0), reused 0 (delta 0), pack-reused 2980
Receiving objects: 100% (2980/2980), 362.27 MiB | 275.00 KiB/s, done.
Resolving deltas: 100% (1769/1769), done.
Checking connectivity… done.
root@ubuntu:~/workspace# git checkout dogsvscats
fatal: Not a git repository (or any of the parent directories): .git
root@ubuntu:~/workspace# cd ncappzoo
apps caffe CONTRIBUTING.md data LICENSE Makefile MAKEFILE_GUIDANCE.md OWNERS README.md tensorflow
root@ubuntu:~/workspace/ncappzoo# git checkout dogsvscats
Branch dogsvscats set up to track remote branch dogsvscats from origin.
Switched to a new branch 'dogsvscats'
apps caffe CONTRIBUTING.md data LICENSE Makefile OWNERS README.md tensorflow
root@ubuntu:~/workspace/ncappzoo# cd apps
benchmarkncs hello_ncs_cpp Makefile README.md street_cam_threaded
birds hello_ncs_py multistick_cpp stream_infer topcoder_example
classifier-gui image-classifier MultiStick_GoogLeNet stream_ty_gn video_face_matcher
dogsvscats live-image-classifier MultiStick_TF_Inception stream_ty_gn_threaded video_objects
gender_age_lbp log-image-classifier rapid-image-classifier street_cam
Hi @Luis_at_Intel ,
I keep forward the trail and meet W/ the make error just as follow:
AUTHORS.txt create-labels.py data README.md
bvlc_googlenet create-lmdb.sh Makefile
root@ubuntu:~/workspace/ncappzoo/apps/dogsvscats# export CAFFE_PATH=/PATH/TO/CAFFE/INSTALL_DIR
sudo apt-get -y install unzip
Reading package lists… Done
Building dependency tree
Reading state information… Done
unzip is already the newest version (6.0-20ubuntu1).
0 upgraded, 0 newly installed, 0 to remove and 346 not upgraded.
sudo pip install pyyaml
DEPRECATION: Python 2.7 will reach the end of its life on January 1st, 2020. Please upgrade your Python as Python 2.7 won't be maintained after that date. A future version of pip will drop support for Python 2.7.
Requirement already satisfied: pyyaml in /usr/local/lib/python2.7/dist-packages (3.13)
replace data/test1/1.jpg? [y]es, [n]o, [A]ll, [N]one, [r]ename:
Creating labels file…
Done. See data/train/train.txt & val.txt
echo "Creating lmdb files…"
Creating lmdb files…
Creating train lmdb…
./create-lmdb.sh: 45: ./create-lmdb.sh: /PATH/TO/CAFFE/INSTALL_DIR/build/tools//convert_imageset: not found
Makefile:30: recipe for target 'lmdb' failed
make: *** [lmdb] Error 127
It seems be caused by the wrong caffe path. Do I need to install the caffe or how could I check if I have install the caffe already ? How could I fix this error? Thank you very much!
Glad my explanation made sense. Since there has been some issues already on your install I may suggest to uninstall and reinstall the NCSDK together with the NCAPPZOO. Just to start fresh, you could do this with
make uninstall. I hope this resolves the issue.
For reference I found this information that could be useful, I don't think this is your case but I don't want to assume that.
Important: Caffe Users: If you are installing on a host machine that already has Caffe installed and its directory is already in the PYTHONPATH environment variable, you must remove the existing Caffe directory from the PYTHONPATH environment variable prior to installing the NCSDK. You will need to manually modify the PYTHONPATH to match your development needs such that it points to the Caffe version installed with the NCSDK when using the NCSDK and it points to other Caffe versions when you wish to use those.
Hi @Luis_at_Intel ,
Thanks for your suggestion. I intall the caffe and re-training the dog vs cat demo. It's running OK now. So I think it need a caffe installed environment. Next I will keep finishing this trail. Additionally one question. The NCSDK not include the tensorflow environment, is this right? I just want confirm this. Thanks again.
That is great news, glad the demo is now working as expected. As for your question "the NCSDK not include the tensorflow environment", can you elaborate more? I don't think it is installed by default, but there are some Tensorflow networks found in the NCAPPZOO that could be useful to get started with. Find more information in the repo here. You can find some Tensorflow references in the NCSDK documentation here as well . Hope these suggestions help answer your question.