def main():
data_path = '../../databases'
PlotsDirectory = '../plots/Week3/task1/'
if not os.path.exists(PlotsDirectory):
os.makedirs(PlotsDirectory)
names = ['highway', 'fall', 'traffic']
estimation_range = [np.array([1050, 1200]), np.array([1460, 1510]), np.array([950, 1000])]
prediction_range = [np.array([1201, 1350]), np.array([1511, 1560]), np.array([1001, 1050])]
#alpha = [{'min': 4, 'max': 20, 'step': 1.5}, {'min': 1, 'max': 10, 'step': 1}, {'min': 1, 'max': 20, 'step': 1.5}]
#ro = [{'min': 1, 'max': 10, 'step': 1}, {'min': 1, 'max': 10, 'step': 1}, {'min': 1, 'max': 10, 'step': 1}]
params = { 'highway': {'alpha': 2.89, 'rho': 0.21},
'fall': {'alpha': 3.2, 'rho': 0.05},
'traffic': {'alpha': 3.55, 'rho': 0.16}}
for i in range(len(names)):
[X_est, y_est] = load_data(data_path, names[i], estimation_range[i], grayscale=True)
[X_pred, y_pred] = load_data(data_path, names[i], prediction_range[i], grayscale=True)
masks = task4(X_est, X_pred, params['traffic']['rho'], params['traffic']['alpha'], True)
masksno = task4(X_est, X_pred, params['traffic']['rho'], params['traffic']['alpha'], False)
print(names[i] + ": F1 score with shadow = " + str(evaluate(masks, y_pred, 'f1')))
print(names[i] + ": F1 score without shadow = " + str(evaluate(masksno, y_pred, 'f1')))
def compute_AUC(X_est, X_pred, y_pred, alpha_range, rho, pixels):
Pr = []
Re = []
for alpha in alpha_range:
#print(alpha)
X_res = task2(X_est, X_pred, rho, alpha, pixels)
Pr.append(evaluate(X_res, y_pred, "precision"))
Re.append(evaluate(X_res, y_pred, "recall"))
return metrics.auc(Re, Pr, True)
metric="precision")
estR_w2 = EstimatorAdaptative(alpha=alpha,
rho=rho[i],
metric="recall")
estP_w2.fit(X_est)
estR_w2.fit(X_est)
X_res_A = task3(X_est, X_pred, rho[i], alpha, True)
X_res_B = task3(X_est, X_pred, rho[i], alpha, False)
X_res_h4, _ = task1(X_est, X_pred, rho[i], alpha, connectivity=4)
X_res_t2 = task2(X_est, X_pred, rho[i], alpha, pixels[i])
X_res_sh_A = task4(X_est, X_pred, rho[i], alpha, True)
X_res_sh_B = task4(X_est, X_pred, rho[i], alpha, False)
Pr_w2.append(estP_w2.score(y_pred, X=X_pred))
Re_w2.append(estR_w2.score(y_pred, X=X_pred))
Pr_h4.append(evaluate(X_res_h4, y_pred, "precision"))
Re_h4.append(evaluate(X_res_h4, y_pred, "recall"))
Pr_A.append(evaluate(X_res_A, y_pred, "precision"))
Re_A.append(evaluate(X_res_A, y_pred, "recall"))
Pr_B.append(evaluate(X_res_B, y_pred, "precision"))
Re_B.append(evaluate(X_res_B, y_pred, "recall"))
Pr_t2.append(evaluate(X_res_t2, y_pred, "precision"))
Re_t2.append(evaluate(X_res_t2, y_pred, "recall"))
Pr_sh_A.append(evaluate(X_res_sh_A, y_pred, "precision"))
Re_sh_A.append(evaluate(X_res_sh_A, y_pred, "recall"))
Pr_sh_B.append(evaluate(X_res_sh_B, y_pred, "precision"))
Re_sh_B.append(evaluate(X_res_sh_B, y_pred, "recall"))
np.save(PlotsDirectory + names[i] + '_Pr_w2.npy', Pr_w2)
np.save(PlotsDirectory + names[i] + '_Re_w2.npy', Re_w2)
np.save(PlotsDirectory + names[i] + '_Pr_h4.npy', Pr_h4)
estimation_range[i],
grayscale=True)
[X_pred, y_pred] = load_data(data_path,
names[i],
prediction_range[i],
grayscale=True)
alpha_range = np.arange(a[i].get('min'), a[i].get('max'), a[i].get('step'))
for idx, alpha in enumerate(alpha_range):
print(str(idx) + "/" + str(len(alpha_range)) + " " + str(alpha))
X_res_h4, _ = task1(X_est, X_pred, rho[i], alpha, connectivity=4)
X_res_h8, _ = task1(X_est, X_pred, rho[i], alpha, connectivity=8)
if doROC:
FPR_h4.append(evaluate(X_res_h4, y_pred, "fpr"))
TPR_h4.append(evaluate(X_res_h4, y_pred, "tpr"))
FPR_h8.append(evaluate(X_res_h8, y_pred, "fpr"))
TPR_h8.append(evaluate(X_res_h8, y_pred, "tpr"))
else:
Pr_h4.append(evaluate(X_res_h4, y_pred, "precision"))
Re_h4.append(evaluate(X_res_h4, y_pred, "recall"))
Pr_h8.append(evaluate(X_res_h8, y_pred, "precision"))
Re_h8.append(evaluate(X_res_h8, y_pred, "recall"))
plt.figure()
if doROC:
line4, = plt.plot(np.array(FPR_h4),
np.array(TPR_h4),
'b',
def main():
data_path = '../../databases'
PlotsDirectory = '../plots/Week3/task3/'
if not os.path.exists(PlotsDirectory):
os.makedirs(PlotsDirectory)
names = ['highway', 'fall', 'traffic']
estimation_range = [
np.array([1050, 1200]),
np.array([1460, 1510]),
np.array([950, 1000])
]
prediction_range = [
np.array([1201, 1350]),
np.array([1511, 1560]),
np.array([1001, 1050])
]
a = [{
'min': 0,
'max': 40,
'step': 1
}, {
'min': 0,
'max': 40,
'step': 1
}, {
'min': 0,
'max': 40,
'step': 1
}]
params = {
'highway': {
'alpha': 7.25,
'rho': 0.6
},
'fall': {
'alpha': 3.2,
'rho': 0.004
},
'traffic': {
'alpha': 10.67,
'rho': 0
}
}
n_pixels = 20
for i in range(len(names)):
#i = 0
[X_est, y_est] = load_data(data_path,
names[i],
estimation_range[i],
grayscale=True)
[X_pred, y_pred] = load_data(data_path,
names[i],
prediction_range[i],
grayscale=True)
mask3 = task3(X_est, X_pred, params[names[i]]['rho'],
params[names[i]]['alpha'], True)
maskno3 = task3(X_est, X_pred, params[names[i]]['rho'],
params[names[i]]['alpha'], False)
print(names[i] + ": F1 score new = " +
str(evaluate(mask3, y_pred, 'f1')))
print(names[i] + ": F1 score past = " +
str(evaluate(maskno3, y_pred, 'f1')))
pr = evaluate(mask3, y_pred, "precision")
re = evaluate(mask3, y_pred, "recall")
pr_no = evaluate(maskno3, y_pred, "precision")
re_no = evaluate(maskno3, y_pred, "recall")