import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from skimage.io import imread
from skimage.feature import hog
from skimage.filters import threshold_otsu



#Citim baza de date pentru antrenare si testare
df = pd.read_csv('eye_data_lab4.csv')
df_t = pd.read_csv('eye_data_lab4_test.csv')

#pandas.DataFrame.nunique finds the unique elements(lines) of a dataset
df['label'].nunique()

#Print dataframe
print(df.head())

# Eliminam coloana label pentru a ramane doar cu caracterisiticile pozelor
features = df.drop('label', axis=1)
label = df['label']

#convertim tipul de date din panda in ndarray
testFeatures = df_t.to_numpy()

#Extragem dimensiunea matricei de caracteristici
n, m = features.shape

# Construiti si rezolvati problema cmmp utilizand comanda lstqr
classifier = np.linalg.lstsq(features, label, rcond=None)[0]

i = input('Introduceti o valoare intre 0 si 3: ')

#Afisarea imaginii testate
img = imread(i +'.jpg')

plt.axis("off")
plt.imshow(img,cmap="gray")
print(img.shape)

thresh = threshold_otsu(img)
binary = img > thresh
testFeatures = hog(binary, orientations=1, pixels_per_cell=(1,1), cells_per_block=(1,1), visualize=False, normalize=True)

# Testati clasificatorul, cu alte cuvinte preziceti test_label
test_label = np.dot(testFeatures, classifier)

if test_label.all() >= 0:
    print('Ochiul este deschis')
else: print('Ochiul este inchis')