def __init__(self, params, original_filename, vocabulary_path=None,
only_eval=False, text_to_eval=None):
Data.__init__(self, original_filename, vocabulary_path, params, text_to_eval)
self.only_eval = only_eval # if the file does not have targets
if text_to_eval is not None:
self.only_eval = True
def load_and_test(model_dir):
a = Data()
model = Model(data=a)
model.load_model(log_model_dir=model_dir)
model.test(test_image_id=50, save_dir=model.test_image_save_dir)
model.test(test_image_id=350, save_dir=model.test_image_save_dir)
for i in range(1, 500):
model.test(i, save_dir=model.test_image_save_dir)
model.end()
def train():
a = Data()
model = Model(data=a)
model.log_config()
model.train()
model.save_model()
# model.load_model(
# log_model_dir='/home/mars/ANN/dls/PatternRecognitionCourseFinalProject/log/6-11-17-21-16/model/model.ckpt-50000')
# model.train()
model.test(test_image_id=50, save_dir=model.test_image_save_dir)
model.test(test_image_id=350, save_dir=model.test_image_save_dir)
model.end()
def removing_objects(data):
sets = [data.train_set, data.test_set]
for i in sets:
# Replacing str values with boolean 0 and 1 values
sex = {'male': 0, 'female': 1}
i.sex = [sex[item] for item in i.sex]
smoker = {'no': 0, 'yes': 1}
i.smoker = [smoker[item] for item in i.smoker]
region = {
'northwest': 1,
'southeast': 2,
'northeast': 3,
'southwest': 4
}
i.region = [region[item] for item in i.region]
return Data(sets[0], sets[1])
import argparse
import matplotlib.pyplot as plt
import numpy as np
from src.data.Data import Data
from src.utils.Config import Config
parser = argparse.ArgumentParser(description="")
parser.add_argument('config')
args = parser.parse_args()
config = Config.from_file(args.config)
data = Data(config.get_with_prefix("data"))
dataset = data.build_val_dataset()
for reference_images, reference_cam_poses, query_images, query_cam_poses, iou, room_ids, pose_transform, full_matching in dataset:
fig = plt.figure()
plt.imshow(np.concatenate((reference_images[0], query_images[0]), axis=1),
extent=[0, data.image_size * 2, data.image_size, 0])
lines = []
def onclick(event):
print('button=%d, x=%d, y=%d, xdata=%f, ydata=%f' %
(event.button, event.x, event.y, event.xdata, event.ydata))
x = int(event.xdata)
y = int(event.ydata)
if 0 <= x < 128 and 0 <= y < 128:
for line in lines:
from src.utils.Config import Config
from src.utils.Inference import Inference
import matplotlib.pyplot as plt
import numpy as np
import imageio
import cv2
parser = argparse.ArgumentParser(description="")
parser.add_argument('config')
parser.add_argument('model_dir')
parser.add_argument('image1')
parser.add_argument('image2')
args = parser.parse_args()
config = Config.from_file(args.config)
data = Data(config.get_with_prefix("data"))
model = ExReNet(config.get_with_prefix("model"), data)
model.load_weights(str(Path(args.model_dir) / "model.h5"))
image1 = imageio.imread(args.image1)
image1 = cv2.resize(image1, (data.image_size, data.image_size))
image2 = imageio.imread(args.image2)
image2 = cv2.resize(image2, (data.image_size, data.image_size))
cam_pose, matched_coordinates, all_dots, matching = model(image1[None] / 255.0,
image2[None] / 255.0,
training=False)
print("Click on the left image to see the matched point in the other image.")
full_matching = np.zeros((32, 32, 2))
def load_drop_empty(train_set_path, test_set_path):
train = pd.read_csv(train_set_path)
test = pd.read_csv(test_set_path)
return Data(train.dropna(), test.dropna())