def _(msg):
if peforth.vm.debug == 55:
peforth.ok('55> ', loc=locals(),
cmd=":> [0] constant loc55 cr") # breakpoint
if msg.user.NickName[:5] == 'AILAB': # 只在 AILAB 工作，過濾掉其他的。
msg.download(msg.fileName)
return 'Attachment: %s received at %s' % (msg.fileName, time.ctime())
def console(msg, cmd):
if cmd:
print(cmd) # already on the remote side, don't need to echo
global nextDelay
nextDelay_msg = '\nNext anti-robot delay time: %i seconds\n' % (
nextDelay)
if peforth.vm.debug == 11:
peforth.ok('11> ', loc=locals(),
cmd=":> [0] constant loc11 cr") # breakpoint
# re-direct the display to peforth screen-buffer
peforth.vm.dictate("display-off")
try:
# peforth.vm.dictate(cmd)
peforth.ok('Console> ',
loc=locals(),
cmd=":> [0] constant console.locals " + cmd + " exit")
except Exception as err:
errmsg = "Failed! : {}".format(err)
peforth.vm.dictate("display-on")
send_chunk(errmsg + nextDelay_msg, msg.user.send)
else:
peforth.vm.dictate("display-on screen-buffer")
screen = peforth.vm.pop()[0]
send_chunk(screen + '\nOK\n' + nextDelay_msg, msg.user.send)
def training_step(i, update_test_data, update_train_data):
# training on batches of 100 images with 100 labels
batch_X, batch_Y = mnist.train.next_batch(100)
# compute training values for visualisation
if update_train_data:
a, c, im, w, b = sess.run([accuracy, cross_entropy, I, allweights, allbiases], feed_dict={X: batch_X, Y_: batch_Y})
datavis.append_training_curves_data(i, a, c)
datavis.append_data_histograms(i, w, b)
datavis.update_image1(im)
print(str(i) + ": accuracy:" + str(a) + " loss: " + str(c))
# compute test values for visualisation
if update_test_data:
a, c, im = sess.run([accuracy, cross_entropy, It], feed_dict={X: mnist.test.images, Y_: mnist.test.labels})
datavis.append_test_curves_data(i, a, c)
datavis.update_image2(im)
print(str(i) + ": ********* epoch " + str(i*100//mnist.train.images.shape[0]+1) + " ********* test accuracy:" + str(a) + " test loss: " + str(c))
# learning_rate_decay
irmax = 0.005
irmin = 0.00001
knee = 1000
ir = irmin + (irmax-irmin)*10**(-i/knee)
# the backpropagation training step
sess.run(train_step, feed_dict={X: batch_X, Y_: batch_Y, learning_rate_decay:ir})
if peforth.vm.debug==22: peforth.ok('',loc=locals(), cmd = "---xray--- marker ---xray--- :> [0] inport i autoexec exit")
def _(msg):
if peforth.vm.debug == 44:
peforth.ok('44> ', loc=locals(),
cmd=":> [0] constant loc44 cr") # breakpoint
if msg.user.NickName[:5] == 'AILAB': # 只在 AILAB 工作，過濾掉其他的。
if msg.isAt:
cmd = msg.text.split(maxsplit=1)[1] # remove the leading @nickName
console(msg, cmd)
def _(msg):
if peforth.vm.debug == 2211:
peforth.ok('2211> ', loc=locals(),
cmd=":> [0] constant loc2211 cr") # breakpoint
# msg.download(msg.fileName) # 放在 working directory 下
# pred = model.classify(image_path=msg.fileName.strip())
# results = model.print_scores(pred=pred,k=10,only_first_name=True)
return predict(msg)
def picture(msg):
if peforth.vm.debug == 55:
peforth.ok(
'55> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr") # breakpoint
if msg.user.NickName == chatroom: # 只在特定的 chatroom 工作，過濾掉其他的。
predict(msg)
def _(msg):
if peforth.vm.debug == 22:
peforth.ok('22> ', loc=locals(), cmd=":> [0] inport cr") # breakpoint
msg.download(msg.fileName)
typeSymbol = {
PICTURE: 'img',
VIDEO: 'vid',
}.get(msg.type, 'fil')
return '@%s@%s' % (typeSymbol, msg.fileName)
def _(msg):
global nextDelay
time.sleep(nextDelay) # Anti-Robot delay
nextDelay = random.choice(range(3, 18))
nextDelay_msg = '\nNext anti-robot delay time: %i seconds\n' % (nextDelay)
if peforth.vm.debug == 55:
peforth.ok('55> ', loc=locals(),
cmd=":> [0] constant loc55 cr") # breakpoint
if msg.user.NickName[:5] == 'AILAB': # 只在 AILAB 工作，過濾掉其他的。
send_chunk(predict(msg) + nextDelay_msg, msg.user.send)
def _(msg):
if peforth.vm.debug == 44:
peforth.ok('44> ', loc=locals(),
cmd=":> [0] constant loc44 cr") # breakpoint
if msg.user.NickName[:5] == 'AILAB': # 只在 AILAB 工作，過濾掉其他的。
if msg.isAt:
time.sleep(nextDelay) # Anti-Robot delay
cmd = (msg.text + "\n").split(
"\n", maxsplit=1)[1] # remove the first line: @nickName ...
console(msg, cmd) # 避免帶有空格的 nickName 惹問題
def _(msg):
if peforth.vm.debug == 33:
peforth.ok('33> ', loc=locals(),
cmd=":> [0] constant loc33 cr") # breakpoint
if msg.user.NickName[:5] == 'AILAB': # 只在 AILAB 工作，過濾掉其他的。
anti_robot_msg = antiRobotSleep()
msg.download('download\\' + msg.fileName)
send_chunk(
'Attachment: %s \nreceived at %s\n' %
(msg.fileName, time.ctime()) + anti_robot_msg, msg.user.send)
def attachment(msg):
if peforth.vm.debug == 33:
peforth.ok(
'33> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr") # breakpoint
if msg.user.NickName == chatroom: # 只在特定的 chatroom 工作，過濾掉其他的。
msg.download('download\\' + msg.fileName)
time.sleep(nextDelay) # Anti-Robot delay
send_chunk(
'Attachment: %s \nreceived at %s\n' %
(msg.fileName, time.ctime()) + nextDelay_msg(), msg.user.send)
def _(msg):
global nextDelay
time.sleep(nextDelay) # Anti-Robot delay
nextDelay = random.choice(range(3, 18))
if peforth.vm.debug == 44:
peforth.ok('44> ', loc=locals(),
cmd=":> [0] constant loc44 cr") # breakpoint
if msg.user.NickName[:5] == 'AILAB': # 只在 AILAB 工作，過濾掉其他的。
if msg.isAt:
cmd = msg.text.split(maxsplit=1)[1] # remove the leading @nickName
console(msg, cmd)
send_chunk('Next anti-robot delay time: %i seconds' % (nextDelay),
msg.user.send)
def _(msg):
global nextDelay
time.sleep(nextDelay) # Anti-Robot delay
nextDelay = random.choice(range(3, 18))
nextDelay_msg = '\nNext anti-robot delay time: %i seconds\n' % (nextDelay)
if peforth.vm.debug == 33:
peforth.ok('33> ', loc=locals(),
cmd=":> [0] constant loc33 cr") # breakpoint
if msg.user.NickName[:5] == 'AILAB': # 只在 AILAB 工作，過濾掉其他的。
msg.download('download\\' + msg.fileName)
send_chunk(
'Attachment: %s \nreceived at %s\n' %
(msg.fileName, time.ctime()) + nextDelay_msg, msg.user.send)
def mfcc_batch_generator(batch_size=10,
source=Source.DIGIT_WAVES,
target=Target.digits):
maybe_download(source, DATA_DIR)
if target == Target.speaker: speakers = get_speakers()
files = os.listdir(path) # all 2402 files include 2 folder
files = [wav for wav in files if wav.endswith(".wav")]
shuffle(files)
shreshold = int(len(files) * test_fraction)
files_test = files[:shreshold]
files_train = files[shreshold:]
def feature_label(wav):
if target == Target.speaker:
label = one_hot_from_item(speaker(wav), speakers)
elif target == Target.digits:
label = dense_to_one_hot(int(wav[0]),
10) # 咱的 case 檔名第一個字母就是 label
elif target == Target.first_letter:
label = dense_to_one_hot((ord(wav[0]) - 48) % 32, 32)
else:
raise Exception("todo : labels for Target!")
wave, sr = librosa.load(path + wav, mono=True)
mfcc = librosa.feature.mfcc(wave, sr)
mfcc = np.pad(mfcc, ((0, 0), (0, 80 - len(mfcc[0]))),
mode='constant',
constant_values=0)
feature = np.array(mfcc)
return feature, label
def getBatch(size, waves):
X = []
Y = []
for i in range(size):
x, y = feature_label(waves[i])
X.append(x)
Y.append(y)
return X, Y
while True:
print("loaded batch of %d train_files, %d test_files" %
(len(files_train), len(files_test)))
shuffle(files_train)
shuffle(files_test)
train_features, train_labels = getBatch(batch_size, files_train)
test_features, test_labels = getBatch(batch_size, files_test)
if peforth.vm.debug == 55: peforth.ok('55x> ', loc=locals())
yield train_features, train_labels, test_features, test_labels
if peforth.vm.debug == 55: peforth.ok('55y> ', loc=locals())
# basic_rnn_seq2seq inputs must be a sequence
pass
def predict(msg):
results = time.ctime() + '\n'
results += 'Google Inception V3 thinks it is:\n'
msg.download('download\\' +
msg.fileName) # 照片放在 working directory/download 下
if peforth.vm.debug == 22:
peforth.ok('22> ', loc=locals(),
cmd=":> [0] constant loc22 cr") # breakpoint
if msg.fileName.strip().lower().endswith((".jpeg", '.jpg', '.png')):
pred = model.classify(image_path=('download\\' + msg.fileName).strip())
results += model.print_scores(pred=pred, k=10, only_first_name=True)
else:
results += 'Ooops! jpeg pictures only, please. {} is not one.\n'.format(
msg.fileName)
return results
def recording(msg):
if peforth.vm.debug == 77:
peforth.ok(
'77> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr") # breakpoint
if msg.user.NickName == chatroom: # 只在特定的 chatroom 工作，過濾掉其他的。
pathname = 'download/' + msg.fileName
msg.download(pathname)
peforth.vm.dictate("sh ffmpeg -i {0} -ar 16000 -ab 256k -ac 1 {0}.wav".
format(pathname) + "\ndrop\n")
result = recognize(audio=pathname + ".wav")
time.sleep(nextDelay) # Anti-Robot delay
send_chunk(
'Mozilla DeepSpeech heard you said: %s \nreceived at %s\n' %
(result, time.ctime()) + nextDelay_msg(), msg.user.send)
def predict(msg):
if peforth.vm.debug == 22:
peforth.ok(
'22> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr") # breakpoint
results = time.ctime() + '\n'
results += 'Google Inception V3 Transfered Learning thinks it is:\n'
pathname = 'download\\' + msg.fileName # 照片放在 working directory/download 下
msg.download(pathname)
# TensorFlow 的 tf.image.decode_bmp/jpen/png/pcm 很差，改用 ffmpeg
peforth.vm.dictate("dos ffmpeg -i {} -y 1.png".format(pathname) +
"\ndrop\n")
results += ai.predict("1.png")
peforth.vm.dictate("dos del {}".format(pathname) + "\ndrop\n")
time.sleep(nextDelay) # Anti-Robot delay
send_chunk(results + nextDelay_msg(), msg.user.send)
def chat(msg):
if peforth.vm.debug == 44:
peforth.ok(
'44> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr") # breakpoint
if msg.user.NickName == chatroom: # 只在特定的 chatroom 工作，過濾掉其他的。
if msg.isAt:
cmd = msg.text.split(
"\n", maxsplit=1)[1] # remove the first line: @nickName ...
console(msg, cmd) # 避免帶有空格的 nickName 惹問題
else:
# Shown on the robot computer
print(time.ctime(msg.CreateTime), end=" ")
for i in msg.User['MemberList']:
if i.UserName == msg.ActualUserName:
print(i.NickName)
print(msg.text)
def chat(msg):
def myConsole(msg):
# 本 chat 當作 peforth 命令處理
cmd = msg.text + '\n' # 保證可以 split()
cmd = cmd.split("\n",
maxsplit=1)[1] # remove the first line: @nickName ...
console(msg, cmd) # 避免帶有空格的 nickName 惹問題
def chat(msg):
# 處裡本 chat, @版主 時當 peforth 命令處理否則只是在 robot PC 上顯示
if msg.isAt:
myConsole(msg)
else:
# Shown on the robot computer
print(time.ctime(msg.CreateTime), end=" ")
for i in msg.User['MemberList']:
if i.UserName == msg.ActualUserName:
print(i.NickName)
print(msg.text)
if peforth.vm.debug == 44:
peforth.ok(
'44> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr") # breakpoint
if msg.user.NickName == chatroom: # 只在特定的 chatroom 工作，過濾掉其他的。
chat(msg)
else:
# 開個後門，否則 Chatroom nickName 改了就控制不到。
# 當 Chatroom nickName 只有部分符合時，並且只針對 PIC 的發言處理。
# 非 @版主 也在 Robot PC 上顯示出來是為了確認 Robot 有反應。
if msg.user.NickName.find(chatroom) != -1:
for i in msg.User['MemberList']:
if peforth.vm.debug == 44:
peforth.ok(
'44a> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr"
) # breakpoint
if (i.UserName == msg.ActualUserName) and (i.NickName in PIC):
chat(msg)
def console(msg, cmd):
if cmd:
print(cmd) # already on the remote side, don't need to echo
if peforth.vm.debug == 11:
peforth.ok('11> ', loc=locals(),
cmd=":> [0] constant loc11 cr") # breakpoint
# re-direct the display to peforth screen-buffer
peforth.vm.dictate("display-off")
try:
peforth.vm.push((locals(), globals(), 'console prompt'))
peforth.vm.dictate(":> [0] to console.locals " + cmd)
# peforth.ok('OK ', loc=locals(),
# cmd=":> [0] to console.locals " + cmd + "\n exit")
except Exception as err:
errmsg = "Failed! : {}\n".format(err)
peforth.vm.dictate("display-on")
send_chunk(errmsg + nextDelay_msg, msg.user.send)
else:
peforth.vm.dictate("display-on screen-buffer")
screen = peforth.vm.pop()[0]
send_chunk(screen + nextDelay_msg, msg.user.send)
def console(msg, cmd):
if cmd:
print(cmd) # already on the remote side, don't need to echo.
peforth.vm.push(msg)
peforth.vm.dictate("to msg") # Availablize msg in peforth interpreter
if peforth.vm.debug == 11:
peforth.ok('11> ', loc=locals(),
cmd=":> [0] to locals cr") # breakpoint
# re-direct the display to peforth screen-buffer
peforth.vm.dictate("display-off")
try:
peforth.vm.push((locals(), globals(), 'console prompt'))
peforth.vm.dictate(":> [0] to locals " + cmd)
except Exception as err:
errmsg = "Failed! : {}".format(err)
peforth.vm.dictate("display-on")
time.sleep(nextDelay) # Anti-Robot delay
send_chunk(errmsg + nextDelay_msg() + "\nOK", msg.user.send)
else:
# Normal cases
peforth.vm.dictate("display-on screen-buffer")
screen = peforth.vm.pop()[0]
time.sleep(nextDelay) # Anti-Robot delay
send_chunk(screen + nextDelay_msg() + "\nOK", msg.user.send)
def attachment(msg):
if peforth.vm.debug == 33:
peforth.ok(
'33> ',
loc=locals(),
cmd=":> [0] to locals locals :> ['msg'] to msg cr") # breakpoint
if msg.user.NickName == chatroom: # 只在特定的 chatroom 工作，過濾掉其他的。
pathname = 'download/' + msg.fileName
msg.download(pathname)
if msg.fileName.lower().endswith(
('.mp3', '.3gpp', '.aac', '.wav', '.wma', '.pcm')):
peforth.vm.dictate(
"sh ffmpeg -i '{0}' -ar 16000 -ab 256k -ac 1 '{0}'.wav".format(
pathname) + "\ndrop\n")
result = recognize(audio=pathname + ".wav")
time.sleep(nextDelay) # Anti-Robot delay
send_chunk(
'Mozilla DeepSpeech heard you said: %s \nreceived at %s\n' %
(result, time.ctime()) + nextDelay_msg(), msg.user.send)
else:
time.sleep(nextDelay) # Anti-Robot delay
send_chunk(
'Attachment: %s \nreceived at %s\n' %
(msg.fileName, time.ctime()) + nextDelay_msg(), msg.user.send)
def mfcc_batch_generator_old(batch_size=10,
source=Source.DIGIT_WAVES,
target=Target.digits):
maybe_download(source, DATA_DIR)
if target == Target.speaker: speakers = get_speakers()
batch_features = []
labels = []
files = os.listdir(path) # 應該是整個 folder 的 .wav 檔 2402 個含兩個 folder
files = [wav for wav in files if wav.endswith(".wav")]
shuffle(files)
files_test = files[:int(len(files) * test_fraction)]
files_train = files[int(len(files) * test_fraction):]
if peforth.vm.debug == 55: peforth.ok('speech_data.py 55a> ', loc=locals())
# batch_size 64
# source is "spoken_numbers_pcm.tar"
# target is Target.digits <enum 'Target'>
while True:
print("loaded batch of %d files" % len(files))
shuffle(files_train)
shuffle(files_test) # 把檔名順序打亂
for wav in files: # wav 是 filename 1_Albert_160.wav 之類
if not wav.endswith(".wav"): continue
wave, sr = librosa.load(path + wav, mono=True)
if target == Target.speaker:
label = one_hot_from_item(speaker(wav), speakers)
elif target == Target.digits:
label = dense_to_one_hot(int(wav[0]),
10) # 咱的 case 檔名第一個字母就是 label
elif target == Target.first_letter:
label = dense_to_one_hot((ord(wav[0]) - 48) % 32, 32)
else:
raise Exception("todo : labels for Target!")
labels.append(label)
mfcc = librosa.feature.mfcc(wave, sr)
# print(np.array(mfcc).shape)
mfcc = np.pad(mfcc, ((0, 0), (0, 80 - len(mfcc[0]))),
mode='constant',
constant_values=0)
batch_features.append(np.array(mfcc))
if peforth.vm.debug == 55: peforth.ok('55b> ', loc=locals())
if len(batch_features) >= batch_size:
# print(np.array(batch_features).shape)
# yield np.array(batch_features), labels
yield batch_features, labels # basic_rnn_seq2seq inputs must be a sequence
if peforth.vm.debug == 66: peforth.ok('66> ', loc=locals())
batch_features = [] # Reset for next batch
labels = []
def _(msg):
if peforth.vm.debug==44: peforth.ok('44> ',loc=locals(),cmd=":> [0] inport cr") # breakpoint
if msg.isAt:
cmd = msg.text.split(maxsplit=1)[1] # remove the leading @nickName
console(msg, cmd)
return data_sets
VALIDATION_SIZE = 2000
local_file = maybe_download(source_data, train_dir)
train_images = extract_images(TRAIN_INDEX, train=True)
train_labels = extract_labels(TRAIN_INDEX, train=True, one_hot=one_hot)
test_images = extract_images(TEST_INDEX, train=False)
test_labels = extract_labels(TEST_INDEX, train=False, one_hot=one_hot)
# train_images = train_images[:VALIDATION_SIZE]
# train_labels = train_labels[:VALIDATION_SIZE:]
# test_images = test_images[VALIDATION_SIZE:]
# test_labels = test_labels[VALIDATION_SIZE:]
data_sets.train = DataSet(train_images, train_labels, load=False)
data_sets.test = DataSet(test_images, test_labels, load=True)
# data_sets.validation = DataSet(validation_images, validation_labels, load=True)
return data_sets
if __name__ == "__main__":
print("downloading speech datasets")
maybe_download(Source.DIGIT_SPECTROS)
maybe_download(Source.DIGIT_WAVES)
maybe_download(Source.NUMBER_IMAGES)
maybe_download(Source.NUMBER_WAVES)
peforth.ok('speech_data_00> ',
loc=locals(),
cmd='''
:> [0] value locals
\ py: vm.debug=55
exit
''')
output_layer = args.output_layer
graph = load_graph(model_file)
t = read_tensor_from_image_file(file_name,
input_height=input_height,
input_width=input_width,
input_mean=input_mean,
input_std=input_std)
input_name = "import/" + input_layer
output_name = "import/" + output_layer
input_operation = graph.get_operation_by_name(input_name)
output_operation = graph.get_operation_by_name(output_name)
with tf.Session(graph=graph) as sess:
start = time.time()
results = sess.run(output_operation.outputs[0],
{input_operation.outputs[0]: t})
end = time.time()
results = np.squeeze(results)
top_k = results.argsort()[-5:][::-1]
labels = load_labels(label_file)
print('\nEvaluation time (1-image): {:.3f}s\n'.format(end - start))
for i in top_k:
print(labels[i], results[i])
peforth.ok('22> ', loc=locals(), cmd=":> [0] value locals cr")
# Initialize debugger peforth
peforth.ok(loc=locals(),
cmd='''
:> [0] value main.locals // ( -- dict ) main locals
\ Check ffmpeg, ffprobe, the needed 'download' directory, and the neural network model
dos ffmpeg -version
[if] cr ." Fatal error! ffmpeg not found which is used to convert pictures." cr bye [then]
dos ffprobe -version
[if] cr ." Fatal error! ffprobe not found which is for pictures' duration." cr bye [then]
dos dir download
[if] cr ." Fatal error! 'download' directory not found." cr bye [then]
dos dir tf_files\models\mobilenet_v1_1.0_224_frozen.tgz
dos dir tf_files\models\mobilenet_v1_0.50_224_frozen.tgz
or \ only one model needed but I have them downloaded already so check them all
[if] cr
." Fatal error! neural network model not found." cr
." tf_files\models\mobilenet_v1_1.0_224_frozen.tgz " cr
." tf_files\models\mobilenet_v1_0.50_224_frozen.tgz" cr
bye [then]
\ define variables
import time constant time // ( -- module )
none value locals // ( -- dict ) at each breakpoint
none value msg // ( -- obj ) itchat dynamic msg package
\ redefine the 'bye' command to logout itchat first
: bye main.locals :> ['itchat'].logout() bye ;
exit \ Don't forget this!!
''')
# Sending message to friend or chatroom depends on the given 'send'
def _(msg):
if peforth.vm.debug==11: peforth.ok('11> ',loc=locals(),cmd=":> [0] inport cr") # breakpoint
send_chunk('%s: %s' % (msg.type, msg.text), msg.user.send)
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import tensorflow as tf
import tensorflowvisu
# -- moved into xray.f -- from tensorflow.examples.tutorials.mnist import input_data as mnist_data
import peforth
peforth.ok(loc=locals(), cmd="include xray.f")
print("Tensorflow version " + tf.__version__)
tf.set_random_seed(0)
# neural network with 1 layer of 10 softmax neurons
#
# · · · · · · · · · · (input data, flattened pixels) X [batch, 784] # 784 = 28 * 28
# \x/x\x/x\x/x\x/x\x/ -- fully connected layer (softmax) W [784, 10] b[10]
# · · · · · · · · Y [batch, 10]
# The model is:
#
# Y = softmax( X * W + b)
# X: matrix for 100 grayscale images of 28x28 pixels, flattened (there are 100 images in a mini-batch)
# W: weight matrix with 784 lines and 10 columns
def _(msg):
if peforth.vm.debug==99: peforth.ok('99> ',loc=locals(),cmd=":> [0] inport cr") # breakpoint
console(msg, msg.Text.strip())
