Skip to content

Latest commit

 

History

History
65 lines (51 loc) · 2.12 KB

README.md

File metadata and controls

65 lines (51 loc) · 2.12 KB

Prerequisite (have openface and torch installed)

Go to openface workspace

$ cd ~/openface/models/openface

Downloading nn4.v1.t7 model

$ wget http://openface-models.storage.cmusatyalab.org/nn4.v1.ascii.t7.xz
$ unxz nn4.v1.ascii.t7.xz

Downloading nn4.small2.v1.t7 model

$ wget http://openface-models.storage.cmusatyalab.org/nn4.small2.v1.ascii.t7.xz
$ unxz nn4.small2.v1.ascii.t7.xz

Convert model from ascii to binary

$ th
th> require 'nn'
th> require 'dpnn'
th> net = torch.load('nn4.v1.ascii.t7', 'ascii')
torch.save('nn4.v1.t7', net)
th> net = torch.load('nn4.small2.v1.ascii.t7', 'ascii')
torch.save('nn4.small2.v1.t7', net)

Solve out of memory problem when installing torch on TK1

modify torch/extra/cutorch/rocks/cutorch-scm-1.rockspec by replacing "-j$(getconf _NPROCESSORS_ONLN)" with "-j1".

Running openface_ros node

$ rosrun openface4ARM openface_ros.py

When interfacing odroid with android

$ rosrun openface4ARM openface_android.py

Start tranning (replace banana with member name. trainingProgress topic = 100 when finish training)

$ rostopic pub /cmdTrainning std_msgs/String "data: 'banana'"

Check capturing progress

$ rostopic echo /capturingProgress

Start recognition

$ rostopic pub -1 /cmdRecognition std_msgs/Bool "data: true"

Check Result

$ rostopic echo /recognitionResults

Reference

Install Openface

Solving Architecture issues

Torch file format

Install Torch on TK1

tk1-torch-install