from keras.layers.core import Reshape from keras.callbacks import ModelCheckpoint from keras.optimizers import Adadelta, RMSprop from sklearn.utils import shuffle import imageio import numpy as np from model import model_final from data_load import load x_train, y_train, x_test, y_test = load() ##### Hyperparameters ############## #################################### shape = (28, 28, 1) epoch = 300 batch_size = 64 CheckDir = "sample/" #################################### model = model_final(shape) ada = Adadelta(lr=5.0, rho=0.95, epsilon=1e-08, decay=0.001) rms = RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.001) model.compile(loss="mean_squared_error", optimizer=rms) model.summary() # import sys # sys.exit()
def choose_rate(call):
data_load.load()
if call.data == 'current rate':
keyboard = telebot.types.InlineKeyboardMarkup()
keyboard.row(
telebot.types.InlineKeyboardButton(text='Рубль',
callback_data='rur_cur'),
telebot.types.InlineKeyboardButton(text='Доллар',
callback_data='usd_cur'))
keyboard.row(
telebot.types.InlineKeyboardButton(text='Евро',
callback_data='eur_cur'),
telebot.types.InlineKeyboardButton(text='Биткойн',
callback_data='btc'))
bot.edit_message_text(chat_id=call.message.chat.id,
message_id=call.message.message_id,
text="Выберите валюту",
reply_markup=keyboard)
elif call.data == 'archive':
keyboard = telebot.types.InlineKeyboardMarkup()
keyboard.row(
telebot.types.InlineKeyboardButton(text='Рубль',
callback_data='rub_arch'),
telebot.types.InlineKeyboardButton(text='Доллар',
callback_data='usd_arch'))
keyboard.add(
telebot.types.InlineKeyboardButton(text='Евро',
callback_data='eur_arch'))
bot.edit_message_text(chat_id=call.message.chat.id,
message_id=call.message.message_id,
text="Выберите валюту",
reply_markup=keyboard)
if call.data == 'rur_cur' or call.data == 'usd_cur' or call.data == 'eur_cur' or call.data == 'btc':
keyboard = telebot.types.InlineKeyboardMarkup()
keyboard.row(
telebot.types.InlineKeyboardButton(text='В начало',
callback_data='start'),
telebot.types.InlineKeyboardButton(text='Назад',
callback_data='back_cur'))
bot.edit_message_text(chat_id=call.message.chat.id,
message_id=call.message.message_id,
text=out_rate(call.data[:3].upper(), 'cur'),
reply_markup=keyboard)
elif call.data == 'rub_arch' or call.data == 'usd_arch' or call.data == 'eur_arch':
keyboard = telebot.types.InlineKeyboardMarkup()
keyboard.row(
telebot.types.InlineKeyboardButton(text='В начало',
callback_data='start'),
telebot.types.InlineKeyboardButton(text='Назад',
callback_data='back_arch'))
bot.edit_message_text(chat_id=call.message.chat.id,
message_id=call.message.message_id,
text=out_rate(call.data[:3].upper(), 'arch'),
reply_markup=keyboard)
if call.data == 'start':
bot.delete_message(chat_id=call.message.chat.id,
message_id=call.message.message_id)
start_message(call.message)
elif call.data == 'back_cur':
call.data = 'current rate'
choose_rate(call)
elif call.data == 'back_arch':
call.data = 'archive'
choose_rate(call)
from NN import NeuralNet
from data_load import load
import matplotlib.pyplot as plt
import numpy as np
# Data load
x, y = load()
shuffle_indices = np.random.permutation((len(y)))
x = x[shuffle_indices]
y = y[shuffle_indices]
train_len = int(len(x) * 0.6)
val_len = int(len(x) * 0.2)
test_len = int(len(x) * 0.2)
x_train = x[:train_len]
y_train = y[:train_len]
x_val = x[train_len:train_len + val_len]
y_val = y[train_len:train_len + val_len]
x_test = x[train_len + val_len:]
y_test = y[train_len + val_len:]
print("x_train len : [ " + str(len(x_train)) + " ]")
print("x_val len : [ " + str(len(x_val)) + " ]")
print("x_test len : [ " + str(len(x_test)) + " ]")
train_size = x_train.shape[0]
batch_size = 20
iters_num = 10000
learning_rate = 0.01
train_loss_list = []
"""
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from data_load import load
BATCH_START = 0
TIME_STEPS = 70 #500 #20
BATCH_SIZE = 50
INPUT_SIZE = 255#1
OUTPUT_SIZE = 1
CELL_SIZE = 50
LR = 0.001
feature = 'feature1_short'
label_file = 'label1_short'
train_data, val_data, test_data, train_label, val_lable, test_label = load()
after = 70
def avaliable_time(after, data):
ava_list = []
startime = np.arange(100, data.shape[0] - 70, 170)
for i in startime:
tmp = list(range(i, i + after))
ava_list.append(tmp)
ava_list = np.asarray(ava_list).flatten()
return ava_list.tolist()
def get_batch(data, label, ava_list):
sess.run(tf.assign(global_step, 0))
if current_epoch > 0:
print("Restoring Checkpoint")
saver.restore(sess, checkpoint_path + "/model-%d" % current_epoch)
print('global step:', sess.run(global_step), 'learning_rate:',
sess.run(learning_rate))
current_epoch += 1
train_iterations = current_epoch * config.net_params[
'total_train'] / batch_size # 训练总数24000
validation_iterations = current_epoch * config.net_params[
'total_validation'] / batch_size
for epoch in range(current_epoch, n_epochs): #
dl.shuffle(1)
for part in range(total_partitions_train):
cam1_container, cam2_container, cam3_container, depth1_container, depth2_container, depth3_container, \
expected_transform_container, velo_pose_container, match_points_container, decalib_mat_container, p_rects, trs = dl.load(part, mode = "train")
for img1, img2, img3, depth_img1, depth_img2, depth_img3, exp_vector, velo_mat, pair_matches, decalib, p_rect, tr in zip(
cam1_container, cam2_container, cam3_container,
depth1_container, depth2_container, depth3_container,
expected_transform_container, velo_pose_container,
match_points_container, decalib_mat_container, p_rects,
trs):
random_disp = np.random.randint(batch_size)
outputs = sess.run(
[
expected_pose, pre_points_2d, predicted_pose, proj_img,
exp_proj_img, train_loss, sift_match_mask,
photometric_loss2, photometric_loss1, vector_loss,
train_step, merge_train, exp_points_2d
],
AP.ToTensor()
])
test_transform = A.Compose([
A.Resize(224, 224, interpolation=1, always_apply=True, p=1),
A.Normalize(mean=mean, std=std),
AP.ToTensor()
])
train_set, test_set = form_data(unzip=True,
data_path=LOCAL_ZIP_NAME,
label_file=LABEL_FILE,
img_dir=IMAGE_DIR,
train_transform=train_transform,
test_transform=test_transform)
trainloader, testloader = load(train_set, test_set, batch_size=128)
use_cuda = torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")
print(device)
if MODEL_NAME == 'mobilenetv2':
print('model:mobilenetv2')
model = torch.hub.load('pytorch/vision', 'mobilenet_v2',
pretrained=True).to(device)
model.classifier[1] = nn.Sequential(
nn.Linear(1280, 256),
nn.ReLU(),
nn.Linear(256, CLASSES),
).to(device)
elif MODEL_NAME == 'resnet18':
tf.summary.scalar('eval_loss', loss)
return tf.estimator.EstimatorSpec(mode=mode,
loss=loss,
eval_metric_ops=eval_metric_ops)
#%%
if __name__ == "__main__":
root_dir = '/home/pohsuanh/Documents/Computer_Vision/HW6'
# Load training and eval data
train_data, eval_data, test_data, gt = data_load.load()
# Flags
TRAIN = False
PREDICT = True
DRAW_SAMPLE = False
# Construct model
if DRAW_SAMPLE == True:
# pic = np.random.randint((test_data['x']).shape[0])
pic = np.random.randint(len(test_data['x']))
from data_load import clear_data, load, binarize_data, train_test_split
from id3 import ID3
from stat import tree_prune_stat
m_data = clear_data(load("./mushroom.txt"))
for i in range(len(m_data)):
f, l = m_data[i][0], m_data[i][-1]
m_data[i][0] = l
m_data[i][-1] = f
m_binary = binarize_data(m_data)
m_train, m_test = train_test_split(m_binary, 0.8)
m_tree = ID3(m_train)
tree_prune_stat(m_tree, m_train, m_test)
# from tensorflow.keras.utils import to_categorical
from sklearn.preprocessing import LabelEncoder, OneHotEncoder
from dice_net import DiceNet
from data_load import load
from tensorflow.compat.v1 import ConfigProto
from tensorflow.compat.v1 import InteractiveSession
config = ConfigProto()
config.gpu_options.allow_growth = True
session = InteractiveSession(config=config)
print("Num GPUs Available: ",
len(tf.config.experimental.list_physical_devices('GPU')))
print("gpu:", tf.test.gpu_device_name())
X, y = load('data')
type_labels = [t[0] for t in y]
val_labels = [v[1] for v in y]
X = np.array(X)
# X = np.expand_dims(X, 3)
# type_labels = np.array(get_type_labels('data'))
# val_labels = np.array(get_value_labels('data'))
label_encoder = LabelEncoder()
type_encoded = label_encoder.fit_transform(type_labels)
val_encoded = label_encoder.fit_transform(val_labels)
onehot_encoder = OneHotEncoder(sparse=False)
type_onehot = onehot_encoder.fit_transform(type_encoded.reshape(-1, 1))
val_onehot = onehot_encoder.fit_transform(val_encoded.reshape(-1, 1))
from data_load import load
from id3 import ID3
from stat import tree_prune_stat
dane = load("./data1.txt", cast_to_int=True)
test = load("./test1.txt", cast_to_int=True)
tree = ID3(dane)
tree_prune_stat(tree, dane, test)