This is a Python implementation of the feature extraction technique used in the 2006 paper https://lvdmaaten.github.io/publications/papers/MuscleCIS_2006.pdf, written by L.J.P. van der Maaten and P.J. Boon.
Example usage:
FileNameHolder_Train = ["~/Coin1", "~/Coin2, "~/Coin3"]
Train_X = []
for FileName in FileNameHolder_Train:
TempCoinHolder = FindCoin(FileName)
HistogramHolder = SplitIntoConcentric(radius = 250, COIN_IMAGE = TempCoinHolder, Center_Y = 250, Center_X = 250)
Train_X.append(HistogramHolder)
Do not change the SplitIntoConcentric() parameters unless you know what you're doing. The FindCoin() function already resizes the image to a 500x500 size, so there is no need to change anything there.
Zero preprocessing is required, the back-end will take care of segmenting the coin, applying the correct filters, and everything else. Bring up an issue if you have any questions/requests.
NO dimensional reduction is done on the returned 5400-dimensional vector. It is highly recommended to use dimensional reduction (the paper reccomend S-PCA to reduce the vector to 200 dimensions).