def trueskill_layer(model_path, images_dir, csv_path):
"""
Compute a trueskill rating for each image of the directory from 100 comparisons predictions.
Save the results in a csv file.
:param model_path: path of the trained comparisons model
:type model_path: str
:param images_dir: images directory path
:type images_dir: str
:param csv_path: path of csv where results are stored
:type csv_path: str
"""
# Variable initialization
print("Loading model ...")
model = load_model(model_path)
images_path = glob.glob(images_dir + "/*jpg")
pictures = []
images = []
# Build list of Image objects
print("Done\nLoading images...")
pbar = ProgressBar()
for img_path in pbar(images_path):
img = Image(img_path, build_array=True)
img.get_info_from_filename()
images.append(img)
pictures.append(img.preprocess_image(224))
pictures = np.array(pictures)
pbar2 = ProgressBar()
print("Done\nComputing trueskill rating for each image ...")
for i in pbar2(range(len(images))):
# Create 100 contenders and predict a comparison for each
contenders = [rd.randint(0, len(images) - 1) for _ in range(100)]
predictions = model.predict([
np.array([pictures[i] for _ in range(100)]),
np.array([pictures[contenders[k]] for k in range(100)])
])
# Update image trueskill rating for each comparison
for j in range(predictions.shape[0]):
if predictions[j][0] > predictions[j][1]:
images[i].trueskill_rating, images[
contenders[j]].trueskill_rating = tr.rate_1vs1(
images[i].trueskill_rating,
images[contenders[j]].trueskill_rating)
else:
images[contenders[j]].trueskill_rating, images[
i].trueskill_rating = tr.rate_1vs1(
images[contenders[j]].trueskill_rating,
images[i].trueskill_rating)
# Save results in csv
print("Done\nSaving results in csv ...")
trueskill_csv(images, csv_path)
print("Done")
def complete_images_table(self, table):
"""
Complete the images table of the database with the relevant values of the directory images.
:param table: SQL table name
:type table: str
"""
# Variable initialization
cmp_double = 0
cmp_img = 0
# Connection to database
conn, cursor = connection_database(self.db_name, self.host, self.user,
self.password, self.local,
self.ssl_ca)
# Get images paths
images = glob.glob(self.img_dir + "/*jpg")
pbar = progressbar.ProgressBar()
for image_path in pbar(images):
image_name = os.path.basename(image_path)
img = Ci(image_path)
# Get image information
lon, lat, k, h, datetime, global_key = img.get_info_from_filename()
date_sql = mapillary_to_sql_date(datetime)
columns = ("image_key", "latitude", "longitude", "datetime",
"filename")
values = (global_key, lat, lon, date_sql, image_name)
try:
insert_into(cursor, table, columns, values)
cmp_img += 1
except mysql.connector.errors.IntegrityError:
# print("Image already in table", image_name)
cmp_double += 1
continue
# Validate queries
commit_query(conn)
# End connection to database
end_connection(conn, cursor)
# Display information
print("Number of images inserted into database: {} ".format(cmp_img))
print("Number of images already inserted in database : {} ".format(
cmp_double))
def ranking_layer(model_path, images_dir, csv_path):
"""
Predict the score of all the images of the directory with the ranking model. Save the results in a csv file.
:param model_path: path of the trained ranking model
:type model_path: str
:param images_dir: images directory path
:type images_dir: str
:param csv_path: path of csv where results are stored
:type csv_path: str
"""
# Variable initialization
images_path = glob.glob(images_dir + "/*jpg")
images = []
# Load scoring model
print("Loading model ...")
model = load_model(model_path)
# Build list of Image objects
print("Done\nMake score prediction for each image ...")
pbar = ProgressBar()
for img_path in pbar(images_path):
img = Image(img_path, build_array=True)
img.get_info_from_filename()
images.append(img)
img.ranking_score = model.predict(np.array([img.preprocess_image(224)]))[0][0]
# Save results in csv
print("Done\nSaving results in csv ...")
ranking_csv(images, csv_path)
# Data
DATA_FOLDER_AUG = abspath("Comparisons_npy/augmented_1")
DATA_FOLDER = abspath("Comparisons_npy")
COMP_CSV = join(DATA_FOLDER, "duels_date.csv")
# Results
TRAIN_RESULTS_KFOLD = abspath("Training_results/Kfold")
TRAIN_RESULTS_SIMPLE = abspath("Training_results/Simple")
# ----------------------------------------------------------------------------------------------------------------------
# Run functions
# ----------------------------------------------------------------------------------------------------------------------
# ------------------------------------- Test Image class ---------------------------------------------------------------
image_snow = Image(IMG_SNOW, build_array=True, build_green_segmentation=True, green_seg_model=SEG_GREEN_MODEL,
build_snow_segmentation=True, snow_seg_model=SEG_SNOW_MODEL)
print(image_snow)
image_snow.is_blurry()
image_snow.is_overexposed()
image_snow.is_under_green_ratio(0.83)
image_snow.is_wrong_exposed()
image_snow.is_snowy()
print(image_snow)
# ------------------------------------- Test Panorama class ------------------------------------------------------------
pano = Panorama(PANO, build_array=True)
pano.create_4_images(SAVE_DIR)
pano.plot_created_images()
# ------------------------------------- Test Download class ------------------------------------------------------------
d = Download(PHOTO_DIR, LAYER_PATH)