Commit 90c935e3f1dae3308b3f8987866a9b8836e6321c

Authored by David Mayerich
1 parent 538df2a2

updates

Showing 2 changed files with 46 additions and 3 deletions   Show diff stats
python/classify.py
... ... @@ -7,6 +7,8 @@ Created on Sun Jul 23 16:04:33 2017
7 7  
8 8 import numpy
9 9 import colorsys
  10 +import sklearn
  11 +import sklearn.metrics
10 12  
11 13 #generate a 2D color class map using a stack of binary class images
12 14 def classcolor2(C):
... ... @@ -39,6 +41,32 @@ def classcolor2(C):
39 41 #input: X x Y x C image giving the probability P(c |x,y)
40 42 #output: X x Y x C binary class image
41 43  
42   -#create an ROC curve calculator
43   -#input: X x Y x C image giving the probability P(c | x,y)
44   -#output: ROC curve
45 44 \ No newline at end of file
  45 +#calculate an ROC curve given a probability image and mask of "True" values
  46 +def image2roc(P, t_vals, mask=[]):
  47 +
  48 + if not P.shape == t_vals.shape:
  49 + print("ERROR: the probability and mask images must be the same shape")
  50 + return
  51 +
  52 + #if a mask image isn't provided, create one for the entire image
  53 + if mask == []:
  54 + mask = numpy.ones(t_vals.shape, dtype=numpy.bool)
  55 +
  56 + #create masks for the positive and negative probability scores
  57 + mask_p = t_vals
  58 + mask_n = mask - mask * t_vals
  59 +
  60 + #calculate the indices for the positive and negative scores
  61 + idx_p = numpy.nonzero(mask_p)
  62 + idx_n = numpy.nonzero(mask_n)
  63 +
  64 + Pp = P[idx_p]
  65 + Pn = P[idx_n]
  66 +
  67 + Lp = numpy.ones((Pp.shape), dtype=numpy.bool)
  68 + Ln = numpy.zeros((Pn.shape), dtype=numpy.bool)
  69 +
  70 + scores = numpy.concatenate((Pp, Pn))
  71 + labels = numpy.concatenate((Lp, Ln))
  72 +
  73 + return sklearn.metrics.roc_curve(labels, scores)
46 74 \ No newline at end of file
... ...
python/envi.py
... ... @@ -254,6 +254,21 @@ class envi:
254 254 p = p + i.shape[0]
255 255 bar.update(l+1)
256 256 return M
  257 +
  258 + def loadband(self, n):
  259 + X = self.header.samples
  260 + Y = self.header.lines
  261 + B = self.header.bands
  262 +
  263 + band = numpy.zeros((Y, X), dtype=self.header.data_type)
  264 + type_bytes = numpy.dtype(self.header.data_type).itemsize
  265 +
  266 + if self.header.interleave == "bsq":
  267 + self.file.seek(n * X * Y * type_bytes)
  268 + self.file.readinto(band)
  269 +
  270 + return band
  271 +
257 272  
258 273 def __del__(self):
259 274 self.file.close()
260 275 \ No newline at end of file
... ...