def main(argv=None):
imgs = data.Preprocess()[8]
# # 取最后一组进行预测
# i = 1350
# image1 = [imgs[i - 3], imgs[i - 2], imgs[i - 1]]
# image2 = [imgs[i - 72], imgs[i - 48], imgs[i - 24]]
# image3 = [imgs[i - 484], imgs[i - 336], imgs[i - 168]]
# # 融合成多通道
# merged1 = cv2.merge([image1[0], image1[1], image1[2]])
# merged1 = merged1.reshape(1, size, size, 3)
# merged2 = cv2.merge([image2[0], image2[1], image2[2]])
# merged2 = merged2.reshape(1, size, size, 3)
# merged3 = cv2.merge([image3[0], image3[1], image3[2]])
# merged3 = merged3.reshape(1, size, size, 3)
# # 单步预测未来一个时刻的数据
# predict_y = evaluate(merged1, merged2, merged3)
# for predict in predict_y:
# print((412640*predict).astype(int))
test_x1, test_x2, test_x3 = [], [], []
for i in range(1350, 1353):
# 取短期、周期、趋势三组件数据
image1 = [imgs[i - 3], imgs[i - 2], imgs[i - 1]]
image2 = [imgs[i - 72], imgs[i - 48], imgs[i - 24]]
image3 = [imgs[i - 484], imgs[i - 336], imgs[i - 168]]
# 融合成多通道
merged1 = cv2.merge([image1[0], image1[1], image1[2]])
merged2 = cv2.merge([image2[0], image2[1], image2[2]])
merged3 = cv2.merge([image3[0], image3[1], image3[2]])
test_x1.append(merged1)
test_x2.append(merged2)
test_x3.append(merged3)
predict_y = evaluate(test_x1, test_x2, test_x3)
# print(predict_y)
print((412640 * predict_y))
def main(argv=None):
train_x1, train_x2, train_x3, test_x1, test_x2, test_x3, train_y, test_y = data.Preprocess(
)[:8]
print('train size:%s, test size:%s' % (len(train_x1), len(test_x1)))
# 图片块，每次取32张图片
batch_size = 32
num_batch = len(train_x1) // batch_size
print()
# 训练网络
train(train_x1, train_x2, train_x3, test_x1, test_x2, test_x3, train_y,
test_y, batch_size, num_batch)
def main(argv=None):
# 内部条件
train_x1, train_x2, train_x3, test_x1, test_x2, test_x3, train_y, test_y = data.Preprocess(
)[:8]
# 外部条件
train_ex, test_ex = data.external()
print('train size:%s, test size:%s' % (len(train_x1), len(test_x1)))
# 图片块，每次取32张图片
batch_size = 32
num_batch = len(train_x1) // batch_size
# 训练网络
train(train_x1, train_x2, train_x3, test_x1, test_x2, test_x3, train_y,
test_y, batch_size, num_batch, train_ex, test_ex)
def __init__(self,
tags,
queue_size=20,
require_shuffle=False,
require_log=False,
is_testset=False,
n_skip_frames=0,
random_num=666,
is_flip=False):
self.is_testset = is_testset
self.shuffled = require_shuffle
self.random_num = random_num
self.preprocess = data.Preprocess()
self.raw_img = Image()
self.raw_tracklet = Tracklet()
self.raw_lidar = Lidar()
# skit some frames
self.tags = [
tag for i, tag in enumerate(tags) if i % (n_skip_frames + 1) == 0
]
self.is_flip = is_flip
if self.shuffled:
self.tags = shuffle(self.tags, random_state=self.random_num)
self.tag_index = 0
self.size = len(self.tags)
self.require_log = require_log
self.flip_axis = 1 # if axis=1, flip from y=0. If axis=0, flip from x=0
self.flip_rate = 2 # if flip_rate is 2, means every two frames
self.cache_size = queue_size
self.loader_need_exit = Value('i', 0)
if use_thread:
self.prepr_data = []
self.lodaer_processing = threading.Thread(target=self.loader)
else:
self.preproc_data_queue = Queue()
self.buffer_blocks = [
Array('h', 41246691) for i in range(queue_size)
]
self.blocks_usage = Array('i', range(queue_size))
self.lodaer_processing = Process(target=self.loader)
self.lodaer_processing.start()
def __init__(self,
dir_path,
dates_to_drivers=None,
indice=None,
cache_num=10,
is_testset=False):
self.dates_to_drivers = dates_to_drivers
self.indice = indice
self.cache_num = cache_num
self.preprocess_path = dir_path
self.is_testset = is_testset
self.preprocess = data.Preprocess()
self.raw_img = Image()
self.raw_tracklet = Tracklet()
self.raw_lidar = Lidar()
# load_file_names is like 1_15_1490991691546439436 for didi or 2012_09_26_0005_00001 for kitti.
if indice is None:
self.load_file_names = self.get_all_load_index(
self.preprocess_path, self.dates_to_drivers, is_testset)
self.tags = self.raw_img.get_tags()
else:
# self.load_file_names = indice
self.load_file_names = self.get_specific_load_index(
indice, self.preprocess_path, self.dates_to_drivers,
is_testset)
self.load_once = True
self.size = len(self.tags)
# self.shuffled_file_names = shuffle(self.load_tags, random_state=1)
# for getting current index in shuffled_file_names
self.batch_start_index = 0
# num_frame_used means how many frames are used in current batch, if all frame are used, load another batch
self.num_frame_used = cache_num
# current batch contents
self.train_rgbs = []
self.train_tops = []
self.train_fronts = []
self.train_gt_labels = []
self.train_gt_boxes3d = []
self.current_batch_file_names = []
def __init__(self,
bags,
tags,
queue_size=20,
require_shuffle=False,
require_log=False,
is_testset=False):
self.is_testset = is_testset
self.shuffled = require_shuffle
self.preprocess = data.Preprocess()
self.raw_img = Image()
self.raw_tracklet = Tracklet()
self.raw_lidar = Lidar()
self.bags = bags
# get all tags
self.tags = tags
if self.shuffled:
self.tags = shuffle(self.tags)
self.tag_index = 0
self.size = len(self.tags)
self.require_log = require_log
self.cache_size = queue_size
self.loader_need_exit = Value('i', 0)
if use_thread:
self.prepr_data = []
self.lodaer_processing = threading.Thread(target=self.loader)
else:
self.preproc_data_queue = Queue()
self.buffer_blocks = [
Array('h', 41246691) for i in range(queue_size)
]
self.blocks_usage = Array('i', range(queue_size))
self.lodaer_processing = Process(target=self.loader)
self.lodaer_processing.start()
o.translation = translation
o.rotation = rotation
o.size = tracklet.size
objects[frames_index.index(i + start_frame)].append(o)
return objects
if __name__ == '__main__':
import data
import net.utility.draw as draw
from sklearn.utils import shuffle
preprocess = data.Preprocess()
raw_img = Image()
raw_tracklet = Tracklet()
raw_lidar = Lidar()
tags = shuffle(raw_tracklet.get_tags())
os.makedirs(os.path.join(config.cfg.LOG_DIR, 'test', 'rgb'), exist_ok=True)
os.makedirs(os.path.join(config.cfg.LOG_DIR, 'test', 'top'), exist_ok=True)
for one_frame_tag in tags:
# load
objs = raw_tracklet.load(one_frame_tag)
rgb = raw_img.load(one_frame_tag)
lidar = raw_lidar.load(one_frame_tag)