def show_lanes(sw, img_warped, img_orig, filename):
img = img_warped.copy()
if sw.left:
fitx_points_warped = np.float32([np.transpose(np.vstack([sw.left.fitx, sw.ploty]))])
fitx_points = cv2.perspectiveTransform(fitx_points_warped, perspective_correction_inv)
left_line_warped = np.int_(fitx_points_warped[0])
left_line = np.int_(fitx_points[0])
n = len(left_line)
for i in range(n - 1):
cv2.line(img_orig, (left_line[i][0], left_line[i][1]), (left_line[i + 1][0], left_line[i + 1][1]),
(0, 255, 0), 5)
cv2.line(img, (left_line_warped[i][0], left_line_warped[i][1]),
(left_line_warped[i + 1][0], left_line_warped[i + 1][1]), (0, 255, 0), 5)
if sw.right:
fitx_points_warped = np.float32([np.transpose(np.vstack([sw.right.fitx, sw.ploty]))])
fitx_points = cv2.perspectiveTransform(fitx_points_warped, perspective_correction_inv)
right_line_warped = np.int_(fitx_points_warped[0])
right_line = np.int_(fitx_points[0])
for i in range(len(right_line) - 1):
cv2.line(img_orig, (right_line[i][0], right_line[i][1]), (right_line[i + 1][0], right_line[i + 1][1]),
(0, 0, 255), 5)
cv2.line(img, (right_line_warped[i][0], right_line_warped[i][1]),
(right_line_warped[i + 1][0], right_line_warped[i + 1][1]), (0, 0, 255), 5)
save_dir(img, "lanes_warped_", filename)
return save_dir(img_orig, "lanes_orig_", filename)
def warp(img, filename):
img_persp = img.copy()
cv2.line(img_persp, perspective_dest[0], perspective_dest[1],
(255, 255, 255), 3)
cv2.line(img_persp, perspective_dest[1], perspective_dest[2],
(255, 255, 255), 3)
cv2.line(img_persp, perspective_dest[2], perspective_dest[3],
(255, 255, 255), 3)
cv2.line(img_persp, perspective_dest[3], perspective_dest[0],
(255, 255, 255), 3)
cv2.line(img_persp, perspective_trapezoid[0], perspective_trapezoid[1],
(0, 192, 0), 3)
cv2.line(img_persp, perspective_trapezoid[1], perspective_trapezoid[2],
(0, 192, 0), 3)
cv2.line(img_persp, perspective_trapezoid[2], perspective_trapezoid[3],
(0, 192, 0), 3)
cv2.line(img_persp, perspective_trapezoid[3], perspective_trapezoid[0],
(0, 192, 0), 3)
save_dir(img_persp, "persp_", filename)
return save_dir(
cv2.warpPerspective(img,
perspective_correction,
warp_size,
flags=cv2.INTER_LANCZOS4), "warp_", filename)
def threshold(channel_threshold, channel_edge, filename):
# Gradient threshold
binary = np.zeros_like(channel_edge)
height = binary.shape[0]
threshold_up = 15
threshold_down = 60
threshold_delta = threshold_down - threshold_up
for y in range(height):
binary_line = binary[y, :]
edge_line = channel_edge[y, :]
threshold_line = threshold_up + threshold_delta * y / height
binary_line[edge_line >= threshold_line] = 255
save_dir(binary, "threshold_edge_only_", filename)
save_dir(channel_threshold, "channel_only_", filename)
binary[(channel_threshold >= 140) & (channel_threshold <= 255)] = 255
binary_threshold = np.zeros_like(channel_threshold)
binary_threshold[(channel_threshold >= 140)
& (channel_threshold <= 255)] = 255
return (save_dir(binary, "threshold_", filename),
save_dir(binary_threshold, "threshold_other", filename))
def edge_detection(channel, filename):
edge_x = cv2.Scharr(channel, cv2.CV_64F, 1, 0) # Edge detection using the Scharr operator
edge_x = np.absolute(edge_x)
return save_dir(np.uint8(255 * edge_x / np.max(edge_x)), "edge_", filename)
CLASS = 1
VALI = 0.1
DROP = 0.5
PATIENCE = 50
BATCH = 128
TEST_DATA = sys.argv[1]
SAVE_DIR = sys.argv[2]
MODEL = './model_bi_lstm.h5'
# load word2vec model
wv_model = w2v.load('./word_vec')
# wv_weight = wv_model.wv.syn0
## load vocab
vocab = dict([(k, v.index) for k, v in wv_model.wv.vocab.items()])
x_setence = utils.load_data(TEST_DATA, file_type = 'test')
x_setence = utils.split_setence(x_setence)
x_setence_id = utils.w2id(x_setence, vocab)
x_test = pad_sequences(x_setence_id)
model = load_model(MODEL)
predicted = model.predict(x_test)
predicted = np.around(predicted)
print(predicted)
utils.save_dir(predicted, SAVE_DIR)
import numpy as np
import sys
from keras.models import load_model
import utils
import models
MEAN = 3.58171208604
TEST_DATA = sys.argv[1]
ANS_PATH = sys.argv[2]
MODEL = './model.h5'
x_test = utils.load_data(TEST_DATA, file_type='test')
model = load_model(MODEL, custom_objects={'rmse': models.rmse})
y_pred = model.predict(np.hsplit(x_test, 2))
print('--y_pred:\n', y_pred)
# y_pred = utils.post_process(y_pred, bias = MEAN)
y_pred = utils.post_process(y_pred, round=False, bias=MEAN)
print('--ans:\n', y_pred)
utils.save_dir(data=y_pred, path=ANS_PATH)
