TrainThresholds.py

import cv2
import numpy as np
from matplotlib import pyplot as plt
from mpl_toolkits.mplot3d import axes3d, Axes3D 

# Save/Load File name for RGB values
filename = 'MaroonWindow.npy'
# Selected training img
imgname = 'Images/InitislSet.png'
# load image
img = cv2.imread(imgname)
# load current training dataset
temp = np.load(filename)

# Select rectangular regions of interest.  
# Be careful not to select bad data or this will make our life hell
samples = cv2.selectROIs('Window',img)
cv2.destroyAllWindows()

# Initialize RGB values
#B =np.array([])
#G =np.array([])
#R =np.array([])
B = temp[0]
G = temp[1]
R = temp[2]

for sample in samples:
	# sample has a starting point, top left corner of box, and a width, and a height
	yS = slice(sample[0],sample[0]+sample[2],1)
	xS = slice(sample[1],sample[1]+sample[3],1)
	# extract sample space
	sampleImg = img[xS,yS,:]


	# Convert the RGB values to 1D arrays for processing

	B = np.append(B,np.ndarray.flatten(sampleImg[0],order='C'))
	G = np.append(G,np.ndarray.flatten(sampleImg[1],order='C'))
	R = np.append(R,np.ndarray.flatten(sampleImg[2],order='C'))


# Save newly added training points to dataset
np.save(filename,np.vstack((B,G,R)))
# Make a 3D ScaTTER Of RGB values
fig = plt.figure()
ax = Axes3D(fig)
#ax = fig.add_subplot(111, projection='3d')
i=0
for i in range(0,R.shape[0]-1):
	ax.scatter(R[i], G[i], B[i],color='k')


ax.set_xlabel('B')
ax.set_ylabel('G')
ax.set_zlabel('R')

plt.show()