I have tried
http://lxbifi21.bifi.unizar.es:5000/

the portal works in the expected way, given probably the trained samples
I have done upload of some pictures to test it out



Deploy on a physical machine
8 CPU cores: Intel(R) Xeon(R) CPU E5472  @ 3.00GHz
RAM: 24GB

nvidia-smi 
Sun Nov 15 11:51:14 2015       
+------------------------------------------------------+                       
| NVIDIA-SMI 352.39     Driver Version: 352.39         |                       
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Quadro K2200        Off  | 0000:0A:00.0     Off |                  N/A |
|  0%   37C    P0     1W /  39W |      9MiB /  4095MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+
modinfo nvidia: version 352.39
kernel: 2.6.32-573.7.1.el6.x86_64

- centos6
- cuda-7.5

1- Download latest caffe from github master branch
$ git clone https://github.com/BVLC/caffe.git

2- Follow install instructions
http://caffe.berkeleyvision.org/installation.html

2.1- Pre-requisites
- CUDA: 7.5
- BLAS: openblas-0.2.14-1.el6.x86_64
        blas-3.2.1-4.el6.x86_64
- Boost: boost-1.41.0-27.el6.x86_64
- Install dependencies: yum install protobuf-devel leveldb-devel snappy-devel opencv-devel boost-devel hdf5-devel
                        yum install gflags-devel lmdb-devel


Working in userland, in my home directory
2.2- Some exports
      $ export PATH=/usr/local/cuda/bin:$PATH
      $ export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH


2.3- At LIP we use mudule to have access to home compiled SW and libs, in particular
I had to 
module load opencv-2.4.10 boost-1.57.0
this gives me python 2.7 and numpy and boost
I had to download and compile glog, as described in http://caffe.berkeleyvision.org/install_yum.html

I have modified the Makefile.config so as to manage to compile it according to the libs and includes
I had on my system, still there where some warnings but this is due to the env I had
about the boost libs, I don't know if this will affect later the running of the SW

First I compiled with
make all

and later with
make distribute

IMPORTANT notice: as stated in the deliverable "make portable" is unknown target
it should be substituted by "make distribute"

I have cloned the repo
git clone https://github.com/jimgoo/caffe-oxford102.git
cd caffe-oxford102/
python bootstrap.py

Downloading takes a lot of time, I will not proceed with it for the current report
meaning I will not run the train of the neural network


Furthermore

I have cloned the 2 repos
git clone https://github.com/EGI-Lifewatch-CC/portuguese-flora.git
git clone https://github.com/EGI-Lifewatch-CC/orchidee.git

I tested
cd portuguese-flora
and after I run the download of the images
./get_images.sh

Downloading Acanthus mollis flower 0 data/images/Acanthus_mollis_flower_0.jpg
http://flora-on.pt/getphoto.php?q=Acanthus+mollis&offset=0&structure=flower
./get_images.sh: line 34: data/images/Acanthus_mollis_flower_0.jpg: No such file or directory
data/images/Acanthus_mollis_flower_0.jpg file is an image

So the script in github tries to download from the wrong URL



General consideration: The description to run the neural-network is confusing
(see my previous comments on the deliverable)
It's dificult to follow the way one could run the chain/workflow to train the NN

A use case to run the full chain/workflow should be clearly described in bulletted list