Mercurial > hg > segmentation

annotate cnmf.py @ 11:915c849b17ea

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
added arg in cnmf script to choose to run standard nmf
author mitian
date Mon, 18 May 2015 17:43:48 +0100
parents 294f66d285af
children c01fcb752221
rev   line source
mi@0 1 """
mi@0 2 C-NMF method for segmentation, modified from here:
mi@0 3
mi@0 4 Nieto, O., Jehan, T., Convex Non-negative Matrix Factorization For Automatic
mi@0 5 Music Structure Identification. Proc. of the 38th IEEE International Conference
mi@0 6 on Acoustics, Speech, and Signal Processing (ICASSP). Vancouver, Canada, 2013.
mi@0 7 """
mi@0 8
mi@0 9 __author__ = "Oriol Nieto"
mi@0 10 __copyright__ = "Copyright 2014, Music and Audio Research Lab (MARL)"
mi@0 11 __license__ = "GPL"
mi@0 12 __version__ = "1.0"
mi@0 13 __email__ = "oriol@nyu.edu"
mi@0 14
mi@0 15 import numpy as np
mi@0 16 import pymf
mi@0 17
mi@0 18 # Local stuff
mi@0 19 from utils import SegUtil
mi@0 20
mitian@7 21 # Algorithm params
mitian@7 22 h = 8 # Size of median filter for features in C-NMF
mitian@7 23 R = 15 # Size of the median filter for the activation matrix C-NMF
mitian@7 24 rank = 4 # Rank of decomposition for the boundaries
mitian@7 25 rank_labels = 6 # Rank of decomposition for the labels
mitian@7 26 R_labels = 6 # Size of the median filter for the labels
mi@0 27
mi@0 28 def cnmf(S, rank, niter=500):
mi@0 29 """(Convex) Non-Negative Matrix Factorization.
mi@0 30
mi@0 31 Parameters
mi@0 32 ----------
mi@0 33 S: np.array(p, N)
mi@0 34 Features matrix. p row features and N column observations.
mi@0 35 rank: int
mi@0 36 Rank of decomposition
mi@0 37 niter: int
mi@0 38 Number of iterations to be used
mi@0 39
mi@0 40 Returns
mi@0 41 -------
mi@0 42 F: np.array
mi@0 43 Cluster matrix (decomposed matrix)
mi@0 44 G: np.array
mi@0 45 Activation matrix (decomposed matrix)
mi@0 46 (s.t. S ~= F * G)
mi@0 47 """
mi@0 48 nmf_mdl = pymf.CNMF(S, num_bases=rank)
mi@0 49 nmf_mdl.factorize(niter=niter)
mi@0 50 F = np.asarray(nmf_mdl.W)
mi@0 51 G = np.asarray(nmf_mdl.H)
mi@0 52 return F, G
mi@0 53
mitian@11 54 def nmf(S, rank, nither=500):
mitian@11 55 nmf_mdl = pymf.NMF(S, num_bases=rank, niter=nither)
mitian@11 56 nmf_mdl.factorize()
mitian@11 57 F = np.asarray(nmf_mdl.W)
mitian@11 58 G = np.asarray(nmf_mdl.H)
mitian@11 59 return F, G
mitian@11 60
mitian@11 61
mi@0 62 def most_frequent(x):
mi@0 63 """Returns the most frequent value in x."""
mi@0 64 return np.argmax(np.bincount(x))
mi@0 65
mi@0 66
mi@0 67 def compute_labels(X, rank, R, bound_idxs, niter=300):
mi@0 68 """Computes the labels using the bounds."""
mi@0 69
mi@0 70 X = X.T
mi@0 71 try:
mi@0 72 F, G = cnmf(X, rank, niter=niter)
mi@0 73 except:
mi@0 74 return [1]
mi@0 75
mi@0 76 label_frames = filter_activation_matrix(G.T, R)
mi@0 77 label_frames = np.asarray(label_frames, dtype=int)
mi@0 78
mi@0 79 # Get labels from the label frames
mi@0 80 labels = []
mi@0 81 bound_inters = zip(bound_idxs[:-1], bound_idxs[1:])
mi@0 82 for bound_inter in bound_inters:
mi@0 83 if bound_inter[1] - bound_inter[0] <= 0:
mi@0 84 labels.append(np.max(label_frames) + 1)
mi@0 85 else:
mi@0 86 labels.append(most_frequent(
mi@0 87 label_frames[bound_inter[0]:bound_inter[1]]))
mi@0 88
mi@0 89 return labels
mi@0 90
mi@0 91
mi@0 92 def filter_activation_matrix(G, R):
mi@0 93 """Filters the activation matrix G, and returns a flattened copy."""
mi@0 94 idx = np.argmax(G, axis=1)
mi@0 95 max_idx = np.arange(G.shape[0])
mi@0 96 max_idx = (max_idx, idx.flatten())
mi@0 97 G[:, :] = 0
mi@0 98 G[max_idx] = idx + 1
mi@0 99 G = np.sum(G, axis=1)
mitian@4 100 G = SegUtil.median_filter(G[:, np.newaxis], R)
mi@0 101 return G.flatten()
mi@0 102
mi@0 103
mitian@11 104 def segmentation(X, rank=4, R=15, h=8, niter=300, CNMF=True):
mi@0 105 """
mi@0 106 Gets the segmentation (boundaries and labels) from the factorization
mi@0 107 matrices.
mi@0 108
mi@0 109 Parameters
mi@0 110 ----------
mi@0 111 X: np.array()
mi@0 112 Features matrix (e.g. chromagram)
mi@0 113 rank: int
mi@0 114 Rank of decomposition
mi@0 115 R: int
mi@0 116 Size of the median filter for activation matrix
mi@0 117 niter: int
mi@0 118 Number of iterations for k-means
mi@0 119 bound_idxs : list
mi@0 120 Use previously found boundaries (None to detect them)
mitian@11 121 CNMF : bool
mitian@11 122 If True, use CNMF; otherwise use NMF
mitian@11 123
mi@0 124 Returns
mi@0 125 -------
mi@0 126 bounds_idx: np.array
mi@0 127 Bound indeces found
mi@0 128 labels: np.array
mi@0 129 Indeces of the labels representing the similarity between segments.
mi@0 130 """
mi@0 131
mi@0 132 # Filter
mitian@1 133 X = SegUtil.median_filter(X, M=h)
mi@0 134 X = X.T
mi@0 135
mi@0 136 # Find non filtered boundaries
mi@0 137 bound_idxs = None
mi@0 138 while True:
mi@0 139 if bound_idxs is None:
mi@0 140 try:
mitian@11 141 if CNMF: F, G = cnmf(X, rank, niter=niter)
mitian@11 142 else: F, G = nmf(X, rank, niter=niter)
mi@0 143 except:
mi@0 144 return np.empty(0), [1]
mi@0 145
mi@0 146 # Filter G
mi@0 147 G = filter_activation_matrix(G.T, R)
mi@0 148 if bound_idxs is None:
mi@0 149 bound_idxs = np.where(np.diff(G) != 0)[0] + 1
mi@0 150
mi@0 151 if len(np.unique(bound_idxs)) <= 2:
mi@0 152 rank += 1
mi@0 153 bound_idxs = None
mi@0 154 else:
mi@0 155 break
mi@0 156
mitian@7 157 return G, bound_idxs