Parsing, Ini matrix

.gitignore

@@ -4,3 +4,9 @@
 /Default
 /build
 /doc/generated
+
+
+/tools/DatasetExtraction/.idea
+/tools/DatasetExtraction/__pycache__
+/tools/DatasetExtraction/test_chords_txt
+/tools/DatasetExtraction/testing.py

tools/DatasetExtraction/chord.py

@@ -0,0 +1,28 @@
+"""
+Simple chord implementation
+
+Notes
+-----
+This implementation tries to follow the references and their implementation
+(e.g., https://github.com/jayg996/BTC-ISMIR19)
+"""
+
+
+class Chord:
+    def __init__(self, start_time, end_time, label):
+        self.start_time = start_time
+        self.end_time = end_time
+        self.label = label
+
+    def __repr__(self):
+        return "({}, {}, {})".format(self.start_time, self.end_time, self.label)
+
+
+class Song:
+    def __init__(self, chords_list):
+        self.chords_list = chords_list
+        try:
+            if chords_list[0].label == 'N':
+                chords_list.pop(0)
+        except:
+            print("Error occured")

tools/DatasetExtraction/dataset.py

@@ -0,0 +1,116 @@
+from abc import ABCMeta, abstractmethod
+from accessify import protected
+from chord import Chord, Song
+import numpy as np
+from glob import glob
+
+class Dataset:
+    #i - song, j - chord
+    song_list = []
+    chord_id = {}
+    dir_path = ""
+    window_size = 100.0
+
+    def __init__(self, dir_path, chord_id, window_size):
+        self.dir_path = dir_path
+        self.chord_id = chord_id
+        self.window_size = window_size
+        self.song_list = self.load()
+
+    @protected
+    def validate(self):
+        pass
+
+    #@protected
+    def load(self):
+        file_list = glob(self.dir_path)
+        for file in file_list:
+            self.song_list.append(Song(self.load_chords(file)))
+        return self.song_list
+
+    @protected
+    def load_chords(self, filename):
+        pass
+
+    #returns vector with probabilities (dtype np array)
+    def get_hmm_matrix_i(self):
+        self.song_list = self.load(self.dir_path)
+        i_matrix = np.array(len(self.chord_id), dtype=float)
+        for song in self.song_list:
+            i_matrix[self.chord_id.get(song.chords_list[
+                                      0].label)] += 1
+        i_matrix = i_matrix / len(self.song_list)
+        return i_matrix
+
+    # returns matrix with probabilities (dtype np array)
+    def get_hmm_matrix_t(self):
+        tran_count = 0.0
+        remainder = 0.0
+        drtn = 0.0
+        w_cnt = 0.0
+        t_matrix = np.array((len(self.chord_id), len(self.chord_id)), dtype=float)
+
+        for song_i in self.song_list:
+            t_matrix[self.chord_id.get(song_i[-1].label)][  # analyze last elem
+                self.chord_id.get(song_i[-1].label)] += (song_i[-1].end_time - song_i[-1].start_time) / self.window_size
+            itr = song_i[:-1]  # remove last element (because use elem+1)
+
+            for j in range(itr):
+                curr_chord = song_i.chords_list[j]
+                drtn = curr_chord.end_time - curr_chord.start_time
+                w_cnt = int(drtn / self.window_size)
+
+                if (remainder > self.window_size / 2):
+                    w_cnt = + 1
+
+                remainder = drtn % self.window_size
+
+                tran_count += w_cnt
+
+                t_matrix[self.chord_id.get(curr_chord.label)][    #self trasition
+                    self.chord_id.get(curr_chord.label)] += w_cnt
+
+                t_matrix[self.chord_id.get(curr_chord.label.label)][    #chord changing transition
+                    self.chord_id.get(song_i.chords_list[j + 1].label)] += 1
+        t_matrix = t_matrix / tran_count
+        return  t_matrix
+
+    def get_hmm_matrix_e(self):
+        pass
+
+
+class Iso_Dataset(Dataset):
+
+    #Load chords from a text file, returns list of Chord instances
+    def load_chords(self, filename):
+        chords_list = []
+        with open(filename, 'r') as f:
+            for line in f:
+                if line:
+                    splits = line.split()
+                    if len(splits) == 3:
+                        s = splits[0]
+                        e = splits[1]
+                        l = self.label_error_modify(splits[2])
+                        chords_list.append(Chord(float(s), float(e), l))
+        return chords_list
+
+    #removes label missprints
+    def label_error_modify(self, label):
+        if label == 'Emin/4':
+            label = 'E:min/4'
+        elif label == 'A7/3':
+            label = 'A:7/3'
+        elif label == 'Bb7/3':
+            label = 'Bb:7/3'
+        elif label == 'Bb7/5':
+            label = 'Bb:7/5'
+        elif label.find(':') == -1:
+            if label.find('min') != -1:
+                label = label[:label.find('min')] + ':' + label[label.find('min'):]
+        return label
+
+
+
+
+