def evaluation(gt_mask, pred, thresh=0.5):
'''
evaluate dice. pred and gt_mask should be (h, w) and pred values are 0-1. if the values are not binary, use thesh to make it binary.
'''
if gt_mask.shape != pred.shape:
print('gt_mask and pred should be the same shape')
return
elif len(gt_mask.shape) > 4:
print('too many rank. only accept 3 or 4')
return
elif len(gt_mask.shape) == 3:
print('squeeze to make rank 2')
gt_mask = np.squeeze(gt_mask)
pred = np.squeeze(pred)
# if the pred is 0-1, make it binary with the thresh
pred = np.where(pred > thresh, 1, 0)
# try:
# Compute Dice
gt = np.greater(gt_mask, 0)
pd = np.greater(pred, 0)
# if the ground truth and prediction have no mask, return 1
if (gt.sum() == 0) and (pd.sum() == 0):
return 1.0
# if the ground truth has no mask but prediction has masks, return 0
elif (gt.sum() == 0) and (pd.sum() > 0):
return 0.0
# if the ground truth has mask but prediction has no masks, ofcause return 0
elif (gt.sum() > 0) and (pd.sum() == 0):
return 0.0
else:
dice = 2 * np.logical_and(pd, gt).sum() / (pd.sum() + gt.sum())
return dice
def calc_Precision(truth, pred, classes):
prec_scores = []
# Iterate over each class
for i in range(classes):
try:
gt = np.equal(truth, i)
pd = np.equal(pred, i)
# Calculate precision
prec = np.logical_and(pd, gt).sum() / pd.sum()
prec_scores.append(prec)
except ZeroDivisionError:
prec_scores.append(0.0)
# Return computed precision scores
return prec_scores
def calc_IoU(truth, pred, classes):
iou_scores = []
# Iterate over each class
for i in range(classes):
try:
gt = np.equal(truth, i)
pd = np.equal(pred, i)
# Calculate iou
iou = np.logical_and(pd, gt).sum() / (pd.sum() + gt.sum() - np.logical_and(pd, gt).sum())
iou_scores.append(iou)
except ZeroDivisionError:
iou_scores.append(0.0)
# Return computed IoU
return iou_scores
def calc_DSC(truth, pred, classes):
dice_scores = []
# Iterate over each class
for i in range(classes):
try:
gt = np.equal(truth, i)
pd = np.equal(pred, i)
# Calculate Dice
dice = 2*np.logical_and(pd, gt).sum() / (pd.sum() + gt.sum())
dice_scores.append(dice)
except ZeroDivisionError:
dice_scores.append(0.0)
# Return computed Dice Similarity Coefficients
return dice_scores
def predict(self, X):
import pandas as pd
import numpy as np
import scipy as sp
from classifier.decisionTree import DecisionTree
pd = []
lab = self.labels[:, np.newaxis]
for clf, wt in zip(self.classifiers, self.eweight):
pd = sum((clf.predict(X) == lab).T * wt)
pd /= self.eweights.sum()
if self.n == 2:
pd[:, 0] *= -1
pd = pd.sum(axis=1)
return self.labels.take(pd > 0, axis=0)
return self.labels.take(np.argmax(pd, axis=1), axis=0)