Skip to content

Latest commit

 

History

History
44 lines (30 loc) · 1.3 KB

README.md

File metadata and controls

44 lines (30 loc) · 1.3 KB
Raw

pipe-classification

Imperial College Individual Project: automatic classification of pipe weld images.

setup

for python to find .so files, need their location to be added to LD_LIBRARY_PATH environment variable. also, you must run this program in a bash terminal. in bashrc, set: export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH

make sure you are in a bash terminal.

import pretrained net

under your options.cfg file, under [train], set f = /path/to/pretrained/net

import DeCAF

clone the repo: https://github.com/UCB-ICSI-Vision-Group/decaf-release/ modify decaf/layers/cpp/Makefile: remove '-Wl' from l3 install dependencies: MPI WITHOUT ROOT SOMEHOW! pip install --install-option="--prefix=$HOME/.local" networkx pip install --install-option="--prefix=$HOME/.local" mpi4py pip install --install-option="--prefix=$HOME/.local" numexpr pip install --install-option="--prefix=$HOME/.local" scikit-image pip install --install-option="--prefix=$HOME/.local" pip install --install-option="--prefix=$HOME/.local" pip install --install-option="--prefix=$HOME/.local" pip install --install-option="--prefix=$HOME/.local"

upgrade boost (if you don't have v1.55): wget http://downloads.sourceforge.net/project/boost/boost/1.55.0/boost_1_55_0.tar.gz