Skip to content

Latest commit

 

History

History
80 lines (55 loc) · 1.74 KB

README_internal.md

File metadata and controls

80 lines (55 loc) · 1.74 KB
Raw

Machine Learning for Object Recognition at Ocado

ensure correct read/write permissions on Group Directory

As per emails of Fidelis/Duncan, add

umask 002

to the end of .bashrc file on your home directory

How to Generate training data

Get the correct files by pulling the branch pavel/pipeline (Once merged this will be changed)

ssh into the GPU: ssh @gpu04.doc.ic.ac.uk

navigate into src/rendering run:

python render_pipeline.py

The generated zip files are in: /vol/project/2017/530/g1753002/render_workspace/final_zip

The render_workspace contains all the necessary data for generation. Once your generation is complete, please copy the zip files out of the folder If somebody tries to generate another set with same name, it will override your files. So please keep the folder ideally completely empty.

How to Train

activate correct CUDA version to link TF to GPU

create a file in your home directory called .bash_profile

with content and save:

if [ -f /vol/cuda/8.0.61-cudnn.7.0.2/setup.sh ]
then
   . /vol/cuda/8.0.61-cudnn.7.0.2/setup.sh
fi

then log out and log in again or restart bash

activating shared ocado venv from anywhere

venv is installed in group folder, run this line in bash

. /vol/project/2017/530/g1753002/ocadovenv/ocadovenv/bin/activate

run

go to Lobster/src/image_retraining/

$ bash run_retrain.bat

this automatically directs retrain.py to the folder

/vol/project/2017/530/g1753002/product-image-dataset

If you want to redirect the training script to a different image folder, run retrain.py as follows:

$ python ./retrain.py --image_dir [some path]

exit venv

$ deactivate

Linter

Run Pep 8 (TDB)

Online version