_miam/_miamG/miam/classification/colorPaletteKmeanDisplay.py

import matplotlib.pyplot as plt
import numpy as np

from . import KmeanDisplay

import miam.aesthetics.Palette as MPAL
import miam.image.ColorSpace as MICS
import miam.pointcloud.PointCloud2D as MPC2D



class colorPaletteKmeanDisplay(KmeanDisplay.KmeanDisplay):
	""" xxx """
	def __init__(self, pcolor= None):

		# colors
		self.colors = ['b', 'g','r','c','m','y','k'] if pcolor == None else pcolor

		# figure and axes
		fig, ax = plt.subplots(2,2) # 1 row x 2 columns
		fig.suptitle('Palette Classification by k-means[color palette distance]')

		wm = plt.get_current_fig_manager()
		wm.window.state('zoomed')

		self.fig = fig
		self.axes = ax

	def plot(self, centroids, assigmentsIdx, iter, convergence,nbSamples):
		# display centroids shape
		self.axes[0,0].cla()
		_range = max([np.amax(np.abs(centroids[:,:,1])),np.amax(np.abs(centroids[:,:,2]))])
		for i,c in enumerate(centroids): 
			# plot palette
			x = c[:,1] # a -> x
			y = c[:,2] # b -> y
			pc = MPC2D.PointCloud2D(x,y)
			xx,yy = MPC2D.PointCloud2D.toXYarray(pc.convexHull())
			self.axes[0,0].plot(xx,yy,self.colors[i%len(self.colors)]+'o--', linewidth=1, markersize=3)
			self.axes[0,0].plot([- _range, _range],[0,0],'k', linewidth=0.1)
			self.axes[0,0].plot([0,0],[-_range,_range],'k', linewidth=0.1)
		self.axes[0,0].set_title("centroids (iter:"+str(iter)+")"+"['b', 'g','r','c','m','y','k']")

		# display image of palette
		self.axes[0,1].cla()
		self.axes[0,1].set_title("palettes("+str(centroids.shape[0])+")")
		palettes = []
		for c in centroids: 
			palettes.append(MPAL.Palette("",c, MICS.ColorSpace.buildLab()))
		img = MPAL.Palette.createImageOfPalettes(palettes)
		img.plot(self.axes[0,1],title=False)

		# distance
		numberOfChange, numberOfRemain, meanDistance = convergence
		self.axes[1,0].cla()
		self.axes[1,0].set_title("mean distance:"+str(meanDistance[-1]/centroids.shape[0]))
		self.axes[1,0].plot(np.asarray(meanDistance)/centroids.shape[0],'r')

		# change
		self.axes[1,1].cla()
		self.axes[1,1].set_title("convergence index (% change)/ absolute:"+str(numberOfChange[-1]))
		self.axes[1,1].plot(100*np.asarray(numberOfChange)/nbSamples,'g')
		# pause
		plt.pause(0.05)