ml.py

from linear import *
from mlp import *
from rbfn import *

init_random()

"""
import numpy as np
import os
from PIL import Image
import matplotlib.pyplot as plt

IMG_SIZE = (8, 8)
PATH = os.path.join("data_large/")
TRAIN = os.path.join(PATH, "train")
classes = os.listdir(TRAIN)


def import_images_and_assign_labels(folder, label, X, Y):
    for file in os.listdir(folder):
        image_path = os.path.join(folder, file)
        im = Image.open(image_path)
        im = im.resize(IMG_SIZE)
        im = im.convert("RGB")
        im_arr = np.array(im)
        im_arr = np.reshape(im_arr, (IMG_SIZE[0] * IMG_SIZE[1] * 3,))
        X.append(im_arr)
        Y.append(label)


def import_dataset():
    X_train, y_train, X_valid, y_valid = [], [], [], []
    labels = np.identity(len(os.listdir(TRAIN)))
    for set_type in ["train", "valid"]:
        for cl, lab in zip(classes, labels):
            if set_type == "train":
                X_set, y_set = X_train, y_train
            else:
                X_set, y_set = X_valid, y_valid
            import_images_and_assign_labels(
                os.path.join(PATH, set_type, cl),
                lab,
                X_set,
                y_set
            )

    return (np.array(X_train) / 255.0, np.array(y_train)), \
           (np.array(X_valid) / 255.0, np.array(y_valid))


(X_train, y_train), (X_valid, y_valid) = import_dataset()

model = create_rbfn_model(len(X_train[0]), 8, len(X_train))
train_rbfn_model(model, X_train, y_train)

for i in X_train[:5]:
    pred = predict_rbfn(model, i, 8)
    print(pred)

destroy_rbfn_model(model)


path = "/Users/redamaizate/Documents/3IABD/Projet-Annuel/ml_python/models/rbf/test1.txt"

save_rbf_model(model,path)
model_load = load_rbf_model(path)
plot_input = [[x,y] for x in range(6) for y in range(6)]
plot_output_colors = ['blue' if predict[0] > predict[1] else 'red' for predict in [predict_rbfn(model_load, coord) for coord in plot_input]]
plt.scatter([p[0] for p in plot_input], [p[1] for p in plot_input], c=plot_output_colors)
plt.show()


X = np.array([
[-3., 3.], [-3., 2.], [-4., 2.], [-4., 3.], [-3.5, 3.5], [4., 1.], [5., 1.], [4., 0.5], [5., 0.5], [4.5, 1.5]
      #[2.5, 4],[2, 3],[3, 3]
])
Y = np.array([
[1., 0.], [1., 0.], [1., 0.], [1., 0.], [1., 0.], [0., 1.], [0., 1.], [0., 1.], [0., 1.], [0., 1.]
      #1,-1,-1
])

num_classes = 2
k = 2
input_dim = 2
expected_output = [y for y in Y]
    #[[1,0] if label >= 0 else [0,1] for label in Y]
model = create_rbfn_model(input_dim, num_classes, k)
train_rbfn_model(model, X, expected_output)
predict = predict_rbfn(model, X[1])

import numpy as np
import os
from PIL import Image
import matplotlib.pyplot as plt

IMG_SIZE = (8, 8)
PATH = os.path.join("data_large/")
TRAIN = os.path.join(PATH, "train")
classes = os.listdir(TRAIN)


def import_images_and_assign_labels(folder, label, X, Y):
    for file in os.listdir(folder):
        image_path = os.path.join(folder, file)
        im = Image.open(image_path)
        im = im.resize(IMG_SIZE)
        im = im.convert("RGB")
        im_arr = np.array(im)
        im_arr = np.reshape(im_arr, (IMG_SIZE[0] * IMG_SIZE[1] * 3,))
        X.append(im_arr)
        Y.append(label)


def import_dataset():
    X_train, y_train, X_valid, y_valid = [], [], [], []
    labels = np.identity(len(os.listdir(TRAIN)))
    for set_type in ["train", "valid"]:
        for cl, lab in zip(classes, labels):
            if set_type == "train":
                X_set, y_set = X_train, y_train
            else:
                X_set, y_set = X_valid, y_valid
            import_images_and_assign_labels(
                os.path.join(PATH, set_type, cl),
                lab,
                X_set,
                y_set
            )

    return (np.array(X_train) / 255.0, np.array(y_train)), \
           (np.array(X_valid) / 255.0, np.array(y_valid))


def accuracy(model):
    true_preds = 0
    total_preds = len(X_train)
    for x, y in zip(X_train, y_train):
        if np.argmax(predict_mlp_model_classification(model, x, input_dim[-1])) == np.argmax(y):
            true_preds += 1
    print(f"Accuracy training: {round((true_preds / total_preds) * 100, 2)}%")

    true_preds = 0
    total_preds = len(X_valid)
    for x, y in zip(X_valid, y_valid):
        if np.argmax(predict_mlp_model_classification(model, x, input_dim[-1])) == np.argmax(y):
            true_preds += 1
    print(f"Accuracy valid: {round((true_preds / total_preds) * 100, 2)}%")


(X_train, y_train), (X_valid, y_valid) = import_dataset()

picture_test = np.random.randint(0, len(X_valid) - 1)

input_dim = [len(X_train[0]), 128, 8]

p_model2 = load_mlp_model("models/mlp/MLP_50000_8x8_8_t_acc-34.56_v_acc-32.36.txt")
test_after = predict_mlp_model_classification(p_model2, X_valid[picture_test], input_dim[-1])
print("After training:", test_after)
print("Class index : ", np.argmax(test_after))
print("Class expected :", np.argmax(y_train[picture_test]))
accuracy(p_model2)
destroy_mlp_model(p_model2)
"""