This application will generate a new dataset to train the Neural Network in https://github.com/albertsanchezf/SimraNN from a dataset that uses SimRa's project dataset format (https://github.com/simra-project/dataset)
Before execution
Depending of the dataset used, a pre processing of the dataset must be done in order to discard rides that have anomalies such as big time spans between samples (i.e. more that X seconds between two consecutives ride samples). There is a program called ReviewDataset (https://github.com/albertsanchezf/ReviewDataset) that will do this preprocessing procedure. It is recommended to preprocess the dataset with a maxMS value of 1000 (or less).
In order to run it:
-
You must change:
- The value of String[] variable 'datasetpaths' located in AdaptDatabase class, to point to your dataset folders.
- The value of String variable 'OUTPUTDATASETPATH' located in AdaptDatabase class, to point the desired output folder.
-
You should modify:
- The value of int[] variable 'DISCARDEDINCIDENTS' located in AdaptDatabase class, if you want to discard any types of incidents
- The value of boolean variable 'BINARYCLASSIFICATION' located in AdaptDatabase class, TRUE if you want a binary classification
- The value of boolean variable 'TERNARYCLASSIFICATION' located in AdaptDatabase class. If you want a ternary classification (TRUE) you MUST indicate the incident type you want to detect in 'ELEMENTINTERNARYCLASSIFICATION'
- The value of boolean variable 'EXTRACTION' located in AdaptDatabase class, if you want to generate the output files
- The value of boolean variable 'USERTAG' located in AdaptDatabase class, if you want to include the user tagged incidents
- The value of integer variable 'MINNUMBEROFREADINGS' located in AdaptDatabase class, to omit data if the number of readings between 3 seconds doesn't reach X samples.
- The value of integer variable 'WINDOWFRAME" located in AdaptDatabase class, to define the window length in miliseconds