UAF-SCNN: a SparseConvNet implementation

About this software

Setup

For the first install,

cd /opt
sudo git clone https://github.com/tbrycekelly/UAF-SCNN.git
sudo chown -R plankline:plankline /opt/UAF-SCNN
sudo chmod 776 -R /opt/UAF-SCNN

To update from github,

cd /opt/UAF-SCNN
git clean -f
git pull

May need to run git config --global --add safe.directory /opt/UAF-SCNN if permissions are not right.

To build SCNN:

cd /opt/UAF-SCNN/build
make clean
make wp2

If there are undefined refences on the make, check that the LIBS line in ./build/Makefile is the output of pkg-config opencv --cflags --libs and inclues -lcublas.

Test it and copy it to final directory (if it works):

./wp2
cp ./wp2 ../scnn

We build the ./scnn executable from the wp2.cpp source code contained in the build subfolder. This should be the starting point for any soruce code edits or modifications to the existing executable.

Quickstart

Options

Command line arguments for ./scnn:

-start NUM  [400]
-stop NUM   [400]
-batchSize NUM  [350]
-train DIR  [Data/plankton/train]
-unl DIR    [data/plankton/test]
-nClasses NUM   [-1]
-exemplarsPerClassPerEpoch NUM  [1000]
-initialLearningRate NUM    [0.003]
-learningRateDecay NUM  [0.01]
-validationSetPercentage NUM    [0]
-cudaDevice NUM
-basename STR   [plankton]

Example command line call (taken from an actual segmentation.py log file):

./scnn -start 324 -stop 324 -unl /tmp/segment/Camera3_VIPF-306-2022-07-21-22-36-51.647 -cD 1

Training a Neural Network

Copy the training dataset into /opt/UAF-SCNN/Data/plankton/train so that images are in subfolders by category, e.g.: /opt/UAF-SCNN/Data/plankton/train/detritus/iamge.jpg

Run classList.sh

cd /opt/UAF-SCNN/Data/plankton
./classList.sh

You may wish to change the minimum sample size required for a taxa to be included by modifying the minN value within classList.sh. Taxa folders with fewer than minN images will not be included in the training.

TODO: Complete this section

Licensing and Use

SparseConvNet is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

SparseConvNet is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.