diff --git a/MLBD.pdf b/MLBD.pdf
index 01aa4f9e744f154143544d16f2445624d75d3d60..74b86eb701abd55e1297f7c9bb027ac72e5ccc14 100644
Binary files a/MLBD.pdf and b/MLBD.pdf differ
diff --git a/rapport.pdf b/rapport.pdf
index f11c304bdf122248d465b48f15d85fe51ae6cb22..9579ea4fe39d22836be1fd211b88475a38fdbd7c 100644
Binary files a/rapport.pdf and b/rapport.pdf differ
diff --git a/scripts/main.py b/scripts/main.py
index 750738b5239437428dd467207fb89e4390bb44e7..c6beea6d8cb8b6a88b751221f96e52a9200332e7 100644
--- a/scripts/main.py
+++ b/scripts/main.py
@@ -127,20 +127,20 @@ unet_model.compile(optimizer='adam', loss=focal_loss(), metrics=['accuracy', 'me
print("model compiled")
-train_image_list = os.listdir('./image_resized/image')
-test_image_list = os.listdir('./image_resized/masks')
+train_image_list = os.listdir('../image_resized/image')
+test_image_list = os.listdir('../image_resized/masks')
number_of_image = sys.argv[1]
if number_of_image == 'all':
- train_image_list = os.listdir('./image_resized/image')
- test_image_list = os.listdir('./image_resized/masks')
+ train_image_list = os.listdir('../image_resized/image')
+ test_image_list = os.listdir('../image_resized/masks')
else:
- train_image_list = os.listdir('./image_resized/image')[:int(number_of_image)]
- test_image_list = os.listdir('./image_resized/masks')[:int(number_of_image)]
+ train_image_list = os.listdir('../image_resized/image')[:int(number_of_image)]
+ test_image_list = os.listdir('../image_resized/masks')[:int(number_of_image)]
# np load the images and masks
-train_image_list = [np.load('./image_resized/image/' + image) for image in train_image_list]
-test_image_list = [np.load('./image_resized/masks/' + image) for image in test_image_list]
+train_image_list = [np.load('../image_resized/image/' + image) for image in train_image_list]
+test_image_list = [np.load('../image_resized/masks/' + image) for image in test_image_list]
model_saver = keras.callbacks.ModelCheckpoint(model_save_path, verbose=1, save_best_only=False, save_freq='epoch')
diff --git a/scripts/prepare.py b/scripts/prepare.py
index a958a2f6565d9cdac5fa2905c397c3c75a0a1e4e..e258b87620738f0ccee948269fa90244e0305cfb 100644
--- a/scripts/prepare.py
+++ b/scripts/prepare.py
@@ -7,8 +7,8 @@ import shutil
import sys
# load dataset in AED/training_images
-train_path = 'AED/training_images'
-test_path = 'AED/test_images'
+train_path = '../AED/training_images'
+test_path = '../AED/test_images'
train_image_list = os.listdir(train_path)
test_image_list = os.listdir(test_path)
@@ -28,8 +28,8 @@ train_masks = []
test_masks = []
# get the point for each image where elephant are located put in a mask
-df_train_csv = pd.read_csv('AED/training_elephants.csv')
-df_test_csv = pd.read_csv('AED/test_elephants.csv')
+df_train_csv = pd.read_csv('../AED/training_elephants.csv')
+df_test_csv = pd.read_csv('../AED/test_elephants.csv')
def getCoordinatesFromImage(imageId):
imageId = imageId.split('.')[0]
@@ -90,25 +90,28 @@ def prepare_data(image_list, path):
# count the number of white pixel in the mask => keep only the image with elephant
nb_white_pixel = np.count_nonzero(mask_crop)
if nb_white_pixel > 0:
- np.save('./image_resized/image/' + str(image) + "_" + str(cpt), image_crop)
- np.save('./image_resized/masks/' + str(image) + "_" + str(cpt), mask_crop)
+ np.save('../image_resized/image/' + str(image) + "_" + str(cpt), image_crop)
+ np.save('../image_resized/masks/' + str(image) + "_" + str(cpt), mask_crop)
if firstTimeForImage:
next_background = True
firstTimeForImage = False
elif next_background:
- np.save('./image_resized/image/' + str(image) + "_" + str(cpt), image_crop)
- np.save('./image_resized/masks/' + str(image) + "_" + str(cpt), mask_crop)
+ np.save('../image_resized/image/' + str(image) + "_" + str(cpt), image_crop)
+ np.save('../image_resized/masks/' + str(image) + "_" + str(cpt), mask_crop)
next_background = False
cpt += 1
print("Image " + str(image_counter) + " done")
image_counter += 1
-shutil.rmtree('./image_resized/image')
-shutil.rmtree('./image_resized/masks')
+if os.path.exists('../image_resized/image'):
+ shutil.rmtree('../image_resized/image')
-os.mkdir('./image_resized/image')
-os.mkdir('./image_resized/masks')
+if os.path.exists('../image_resized/masks'):
+ shutil.rmtree('../image_resized/masks')
+
+os.mkdir('../image_resized/image')
+os.mkdir('../image_resized/masks')
# prepare data
diff --git a/scripts/pres.py b/scripts/pres.py
index f9f5b07096ebc9d94477af4ded1f6e8488ed1a95..f9eb4a3fa2eae0a13a33f6c2f8e7e3e1e1898ddb 100644
--- a/scripts/pres.py
+++ b/scripts/pres.py
@@ -9,7 +9,7 @@ import cv2
# plot the image
-train_path = 'AED/training_images'
+train_path = '../AED/training_images'
train_image_list = os.listdir(train_path)
for image in train_image_list:
diff --git a/scripts/testModel.py b/scripts/testModel.py
index 87f310cc3b22ec156106554f1c963e6bd3f72c57..41b931874b6d31e90e5fc1007bed65d17122dde6 100644
--- a/scripts/testModel.py
+++ b/scripts/testModel.py
@@ -16,18 +16,18 @@ def focal_loss(gamma=2., alpha=.25):
# model = keras.models.load_model('model_1000')
-model = keras.models.load_model('model_1000', custom_objects={'focal_loss_fixed': focal_loss()})
+model = keras.models.load_model('../model_1000', custom_objects={'focal_loss_fixed': focal_loss()})
# test the model on an image and show the result
# open the
# load all image with numpy
-image_list = os.listdir("./image_resized/image")
-mask_list = os.listdir("./image_resized/masks")
+image_list = os.listdir("../image_resized/image")
+mask_list = os.listdir("../image_resized/masks")
for i in range(len(image_list)):
- image_list[i] = np.load(os.path.join("./image_resized/image", image_list[i]))
- mask_list[i] = np.load(os.path.join("./image_resized/masks", mask_list[i]))
+ image_list[i] = np.load(os.path.join("../image_resized/image", image_list[i]))
+ mask_list[i] = np.load(os.path.join("../image_resized/masks", mask_list[i]))
mask = model.predict(np.expand_dims(image_list[i], axis=0))