Relevant files from my MSc project

Steps to run a simulation:

I have not tested whether more steps are required to successfully run a simulation. Let me know if you encounter any issues: felipeduque@riseup.net

I have only tested under Arch Linux operating system. It should run fine under other Linux flavors though.

numpy Pillow (PIL) scipy multiprocessing itertools caffe opencv (cv2)

The student version is free and GPL'd. Download it here.

Unfortunately, third-party EdgeBoxes was written in Matlab. Fortunately, you only need to download the runtime environment (which is free) to run the compiled code. Download it here. The code was compiled in Matlab 2014b.

You can do that by

export PYTHONPATH=/path/to/lib:/path/to/caffe

You can make it persistent by adding it to your .bashrc file.

You can run it by

./run\_my\_edgeboxes.sh ~/matlab ../../code/imgs/temp/image_file_name_bbs.txt models/forest/modelBsds.mat

where ~/matlab is the Matlab runtime environment installation directory. This application periodically checks for changes in the .txt file. The main application, code/code\_10.py, is responsible for making the changes. Thus the application must be running at the same time as the main application.

You must open V-REP inside code. The suggested scene is in code/scenes.

Start the simulation in V-REP.

You can run the main application by

python code\_10.py sgng14.npy dicionary14.npy 0.1

if you want a pre-trained robot. If you want a naive one, run it by

python code\_10.py 0.1