My 4th year Honours project.
The project adapts the decision algorithm developed by Matthew A. Pugh on a massive parallel scale.
https://github.com/mattalexpugh
This aims to remove a large amount of False Positive fish detections in the Fish4Knowledge (F4K) dataset, without losing too many True Positives.
http://groups.inf.ed.ac.uk/f4k/
run______.py - mpi script for each stage of processing.
f4klib.py - contains necessary functions for extraction and processing.
f4klib2.py - contains optional functions for plotting and statistics.
old-processing-scripts\ - contains outdated mpi processing scripts.
old-scripts\ - contains outdated I/O processing scripts for parsing data.
f4k notebook.ipynb - Main experiment notebook.
ML.ipynb and GroundTruth.ipynb - Ground Truthing and Machine Learning experiments.
SVM parameter find.ipynb - Support Vector Machine experiments.
Job partition stuff.ipynb - For generating files necessary for extractions.
final cleaning.ipynb - Voting Strategy experiments.
dissplots.ipynb - Plotting for Dissertation.
newgt\ - Additional Ground Truth Set
mpi\ - mpi-master-slave library.
lua\ - `Neural Network developed by Pugh.
benchmarking\ - Performance testing.
workspace\ - Miscellaneous files needed for extraction.