def process_batch(self, batch_sample):
steering_angle = np.float32(batch_sample[3])
images, steering_angles = [], []
for image_path_index in range(3):
image_name = batch_sample[image_path_index].split('/')[-1]
image = cv2.imread(self.image_path + image_name)
rgb_image = utils.bgr2rgb(image)
resized = utils.crop_and_resize(rgb_image)
images.append(resized)
if image_path_index == 1:
steering_angles.append(steering_angle + self.correction_factor)
elif image_path_index == 2:
steering_angles.append(steering_angle - self.correction_factor)
else:
steering_angles.append(steering_angle)
if image_path_index == 0:
flipped_center_image = utils.flipimg(resized)
images.append(flipped_center_image)
steering_angles.append(-steering_angle)
return images, steering_angles
def telemetry(sid, data):
if data:
# The current steering angle of the car
steering_angle = data["steering_angle"]
# The current throttle of the car
throttle = data["throttle"]
# The current speed of the car
speed = data["speed"]
# The current image from the center camera of the car
imgString = data["image"]
image = Image.open(BytesIO(base64.b64decode(imgString)))
image_array = np.asarray(image)
image_array = crop_and_resize(image_array)
steering_angle = float(
model.predict(image_array[None, :, :, :], batch_size=1))
throttle = 0.2
print(steering_angle, throttle)
send_control(steering_angle, throttle)
# save frame
if args.image_folder != '':
timestamp = datetime.utcnow().strftime('%Y_%m_%d_%H_%M_%S_%f')[:-3]
image_filename = os.path.join(args.image_folder, timestamp)
image.save('{}.jpg'.format(image_filename))
else:
# NOTE: DON'T EDIT THIS.
sio.emit('manual', data={}, skip_sid=True)
def __data_generation(self, list_IDs):
exemplars = np.empty((self.batch_size, 127, 127, self.n_channels))
instances = np.empty((self.batch_size, 255, 255, self.n_channels))
bboxes = []
cats = []
for i, ID in enumerate(list_IDs):
# print (list_IDs)
# exit()
instance_img_data = self.imgs_to_anns[ID]
base_name = os.path.basename(
os.path.normpath(instance_img_data["file_path"]))
base_name = int(base_name.split(".")[0])
min_ = base_name - 100
max_ = base_name + 100
if min_ < 0:
min_ = 0
if max_ > instance_img_data["total"]:
max_ = instance_img_data["total"]
ex_file_name = ""
while ex_file_name not in self.imgs_to_anns:
ex_idx = random.randint(
min_, max_
) #random.choice(self.data[instance_img_data["category_id"]][instance_img_data["video_id"]])
ex_idx = str(ex_idx).zfill(8) + ".jpg"
ex_file_name = instance_img_data["file_path"]
ex_file_name = ex_file_name[:-12]
ex_file_name += ex_idx
# print (self.data[instance_img_data["category_id"]][instance_img_data["video_id"]])
# exit(
# print (self.imgs_to_anns.keys())
exemplar_img_data = self.imgs_to_anns[ex_file_name]
bbox = exemplar_img_data["bbox"]
x, y, w, h = bbox
box = [y, x, h, w]
# print (instance_img_data)
# print (exemplar_img_data)
cats.append(instance_img_data["category_name"])
instance_img = np.array(cv2.imread(instance_img_data["file_path"]))
exemplar_img = np.array(cv2.imread(exemplar_img_data["file_path"]))
x_sz, z_sz = calculate_x_z_sz(box)
instance_img = cv2.resize(instance_img, (255, 255))
pad = 3000
color = [0, 0, 0] #np.mean(exemplar_img, (0, 1))
# exemplar_img = cv2.copyMakeBorder( exemplar_img, pad, pad, pad, pad, cv2.BORDER_CONSTANT, value = color)
# exemplar_img = np.pad(exemplar_img, [(pad, pad), (pad, pad), (0,0)], mode="constant", constant_values=[color, color])
exemplar_img = crop_and_resize(exemplar_img,
box,
x_sz,
out_size=127,
border_value=[0, 0, 0])
# plt.imshow(exemplar_img)
# plt.show()
instance_img = self._image_augment(instance_img)
exemplar_img = self._image_augment(exemplar_img)
instances[i] = instance_img
exemplars[i] = exemplar_img
# positive_label_pixel_radius = 16 # distance from center of target patch still considered a 'positive' match
# response_size = 17
# response_stride = 6.0
# data_size = len(instances)
# label = make_label(response_size, positive_label_pixel_radius / response_stride)
# labels = np.empty((data_size,) + label.shape)
# labels[:] = label
labels = np.array(
list(self.c_labels)
) #construct_batch_gt_score_maps((17, 17), 6, len(instances))
return [instances, exemplars], [labels], cats
box2 = [y, x, h, w]
box2 = np.array(
[box2[1] + (box2[3]) / 2, box2[0] + (box2[2]) / 2, box2[3], box2[2]],
dtype=np.float32)
pos_x, pos_y = box2[:2]
print(image)
image = np.array(cv2.imread(image))
copy_img = np.array(image)
copy_img2 = np.array(image)
copy_img3 = np.array(image)
copy_img = cv2.resize(copy_img, (255, 255))
exemplar = [
crop_and_resize(exemplar, [y, x, h, w],
x_sz * f,
out_size=127,
border_value=[0, 0, 0]) for f in scale_factors
]
# for e in exemplar:
# plt.imshow(np.array(exemplar[0]).astype(np.uint8))
# plt.show()
exemplar = np.stack(exemplar, axis=0)
print(exemplar.shape)
scores = m.predict([np.array([copy_img] * 3), exemplar])
print(scores.shape)
import cv2
import utils
img_left_original = cv2.imread(
"./MyData06/IMG/left_2019_08_11_15_35_48_328.jpg")
img_center_original = cv2.imread(
"./MyData06/IMG/center_2019_08_11_15_35_48_328.jpg")
img_right_original = cv2.imread(
"./MyData06/IMG/right_2019_08_11_15_35_48_328.jpg")
img_l_o_rgb = utils.bgr2rgb(img_left_original)
img_c_o_rgb = utils.bgr2rgb(img_center_original)
img_r_o_rgb = utils.bgr2rgb(img_right_original)
img_left_cropped = utils.crop_and_resize(img_l_o_rgb)
img_center_cropped = utils.crop_and_resize(img_c_o_rgb)
img_right_cropped = utils.crop_and_resize(img_r_o_rgb)
img_left_cropped = cv2.cvtColor(img_left_cropped, cv2.COLOR_RGB2BGR)
img_center_cropped = cv2.cvtColor(img_center_cropped, cv2.COLOR_RGB2BGR)
img_right_cropped = cv2.cvtColor(img_right_cropped, cv2.COLOR_RGB2BGR)
cv2.imwrite("./images/img_left_cropped.jpg", img_left_cropped)
cv2.imwrite("./images/img_center_cropped.jpg", img_center_cropped)
cv2.imwrite("./images/img_right_cropped.jpg", img_right_cropped)
img_left_flipped = utils.flipimg(img_left_cropped)
img_center_flipped = utils.flipimg(img_center_cropped)
img_right_flipped = utils.flipimg(img_right_cropped)
cv2.imwrite("./images/img_left_flipped.jpg", img_left_flipped)
cv2.imwrite("./images/img_center_flipped.jpg", img_center_flipped)
cv2.imwrite("./images/img_right_flipped.jpg", img_right_flipped)