Introduction: Project aims to recognize person with low-resolution camera
- MRSE as key point detection
- Discrete cosine transform as descriptor
- Naive Bayes classifier as recognition
The steps are based on the following paper https://pdfs.semanticscholar.org/f376/711bbe5cc71089e0c74982cafed5f6bdae79.pdf?_ga=1.49846566.1631971690.1490113648
Data Source: http://projects.asl.ethz.ch/datasets/doku.php?id=ir%3Airicra2014 http://www.polymtl.ca/litiv/en/vid/ http://vcipl-okstate.org/pbvs/bench/
The used training data and testing data are zipped as Training.zip and Testing.zip
Usage: Change paths in these two files
TestTraining.m
path_root = 'C:\Users\Kai\Desktop\Infrared-Camera-Person-Detection\Training';
path_training_positives = fullfile(path_root,'\positives');
path_training_negatives = fullfile(path_root,'\negatives');
TestUseClassifier.m
path_root = 'C:\Users\Kai\Desktop\Infrared-Camera-Person-Detection\Testing';
files = dir(fullfile(path_root,'*.jpg'));
Methods TestUseClassifier TrainNaiveBayesClassifier are general version which not require Parallel Computing Toolbox TestUseClassifierM, TrainNaiveBayesClassifierP are parallelize version which require Parallel Computing Toolbox, it is still in develope prcoess and not ready to use due to race condition is not well tested.