CERTH Garment Classifier Package

GARMENT TYPE INSTALLATION GUIDE

1)Run the commands to install Torch globally:

sudo apt-get install curl
curl -sk https://raw.githubusercontent.com/torch/ezinstall/master/install-deps | bash
curl -sk https://raw.githubusercontent.com/torch/ezinstall/master/install-luajit+torch | bash
sudo luarocks install env

In case you need additional details see detailed installation instructions for Torch7 at
https://github.com/torch/torch7/wiki/Cheatsheet

2) Unzip the attached file (deep_rec.zip) into .ros/certh_data folder

3) In order to test Torch7 installation and attached LUA files :

$ cd
$ cd .ros/certh_data/deep_rec/RecGEOunf
$ th    

At this point Torch7 terminal should be open
Into the Torch7 terminal type
dofile 'CNN_Type_Recognition_Geo.lua' 

Expected output :

9.9993e-01  7.4181e-06  5.5122e-05
[torch.DoubleTensor of dimension 1x3]

To exit Torch terminal type "exit"

4) In order to test garment type classifier (be sure that you have the latest code of the repository) , type into a terminal

rosrun certh_libs test_DeepRec.cpp 

Expected output:

 2.7130e-06  1.2855e-04  1.8698e-01  6.2630e-02  7.5023e-01
[torch.DoubleTensor of dimension 1x5]

 For class 1 Probability is : 0.750235
 For class 4 Probability is : 0.186978
 For class 3 Probability is : 0.0626297
 For class 2 Probability is : 0.000128554
 For class 0 Probability is : 2.71299e-06

SKIN RECOGNITION INSTALLATION GUIDE

deep_rec.zip (29.7 MB) Andreas Kargakos, 01/14/2015 15:45