This is a term project during my course Pattern Recognition at Sejong University (Spring 2019). The aim of this project is to apply different methods and techniques that we had acquired throught the semester to the real world problem. For this purpose, we build a robust system which correctly identifies whether the prostate tissue sample is malignant or benign.
The dataset was provided by the professor. It contains 400 images of prostate tissue samples stained with hematoxylin and eosin (H&E). 300 images had been named in the format label+id with prefixes, being either 'b' or 'c'(cancer). These labeled images further are divided into two sets: train set with 200 images (100 Benign, 100 Cancer) and validation set with 100 images (50 Benign, 50 Cancer). Remaining unlabeled 100 images are test set.