Code for finetuning AlexNet in TensorFlow >= 1.2rc0
Home Page: https://kratzert.github.io/2017/02/24/finetuning-alexnet-with-tensorflow.html
License: BSD 3-Clause "New" or "Revised" License
I'm Frederik Kratzert, Researcher @ Google working in the Flood Forecasting Team. This here is my private GitHub account, were I maintain my open source projects and publish code related to any kind of research articles.
Most of my research is dedicated towards solving applications in environmental sciences (mainly hydrology) with machine learning.
Hi,
I tried reproducing the example keeping it as simple as possible. I have train.txt as:
images/cat1.png 0
images/cat2.png 0
images/cat3.png 0
images/dog1.png 1
images/dog2.png 1
images/dog3.png 1
And test.txt as:
images/cat4.png 0
images/dog4.png 1
when i run the finetune.py i get this error:
Traceback (most recent call last):
File "finetune.py", line 163, in
test_acc /= test_count
ZeroDivisionError: float division by zero
Tried debugging the error and found that val_batches_per_epoch is 0 and the inner body isn't executing.
Using the new ImageDataGenerator (June 15 update), I get lower accuracy than the old ImageDataGenerator (using cv2).
I have changed the new ImageDataGenerator very slightly (tf.image.decode_png => tf.image.decode_jpeg, VGG_MEAN => IMAGENET_MEAN), but still get significant difference in accuracy. To benchmark the two datagenerators, I am loading in the BVLC AlexNet weights and validating on the ImageNet validation set without shuffle.
With the old cv2 ImageDataGenerator, I get 55.72% Top-1 Accuracy and 79.08% Top-5 Accuracy. With the new TF ImageDataGenerator, I get 48.58% Top-1 Accuracy and 73.21% Top-5 Accuracy.
I suspect this has to do with how TF loads in images as opposed to CV2. Here is what the two different image processing steps look like:
Old CV2 ImageDataGenerator:
def next_batch(self, batch_size):
"""
This function gets the next n ( = batch_size) images from the path list
and labels and loads the images into them into memory
"""
# Get next batch of image (path) and labels
paths = self.images[self.pointer:self.pointer + batch_size]
labels = self.labels[self.pointer:self.pointer + batch_size]
#update pointer
self.pointer += batch_size
# Read images
images = np.ndarray([batch_size, self.scale_size[0], self.scale_size[1], 3])
for i in range(len(paths)):
img = cv2.imread(paths[i])
#flip image at random if flag is selected
if self.horizontal_flip and np.random.random() < 0.5:
img = cv2.flip(img, 1)
#rescale image
img = cv2.resize(img, (self.scale_size[0], self.scale_size[1]))
img = img.astype(np.float32)
#subtract mean, which is np.array([104., 117., 124.])
img -= self.mean
images[i] = img
# Expand labels to one hot encoding
one_hot_labels = np.zeros((batch_size, self.n_classes))
for i in range(len(labels)):
one_hot_labels[i][labels[i]] = 1
#return array of images and labels
return images, one_hot_labels
New TF ImageDataGenerator:
def _parse_function_inference(self, filename, label):
"""Input parser for samples of the validation/test set."""
# convert label number into one-hot-encoding
one_hot = tf.one_hot(label, self.num_classes)
# load and preprocess the image
img_string = tf.read_file(filename)
img_decoded = tf.image.decode_jpeg(img_string, channels=3)
img_resized = tf.image.resize_images(img_decoded, [227, 227])
IMAGENET_MEAN = tf.constant([104., 117., 124.], dtype=tf.float32)
img_float = tf.to_float(img_resized)
# RGB -> BGR
img_bgr = img_float[:, :, ::-1]
img_centered = tf.subtract(img_bgr, IMAGENET_MEAN)
return img_centered, one_hot
I believe I am using the CV2 Datagenerator correctly as I can successfully finetune and call train_generator.next_batch(batch_size) within the training and validation loops with no issue. I built off of the finetune.py file you provided for the new TF Datagenerator (I use cpu device cpu:0 and then run the datagenerator init ops), and then get batches within the training/validation loops with img_batch, label_batch = sess.run(next_batch).
Any advice? Thanks a lot for providing and maintaining this proejct. Besides for this issue, it was super easy to work with and modify!
Hello, I've never worked with tensorflow before and I cannot really understand how to see some of the output. I am trying to print the the values that fc7 outputs for every image in the testing set. Is there any way I can do this?
I'm getting this error. I already researched and can not find a solution. I'm running on Windows and have already changed the position of the parameters and even then it did not work. Would you help me?
Hello , I try to use your code in my data ,and I create the txt file, by the way , I want to know the format of the txt ,in my txt file , the first column is the photo directory ，and the second is the label . Does this correct? Besides this ,I run the program and it told me that :No such file or directory: 'bvlc_alexnet.npy'
I guess it was a weight file which has been trained before ,but I don't find it ,how should I do ?
Many thanks!
hi, thanks for your sharing.
I am using your code to finetune alexnet to classify images into blur or clear. When I run finetune.py with my 'train.txt' and 'val.txt', after dozens of train_batches, I get this error 👎 tensorflow.python.framework.errors_impl.InvalidArgumentError: Input shape axis 0 must equal 4, got shape [5]
[[Node: unstack_1 = UnpackT=DT_INT32, axis=0, num=4]]
[[Node: IteratorGetNext = IteratorGetNextoutput_shapes=[[?,227,227,3], [?,2]], output_types=[DT_FLOAT, DT_FLOAT], _device="/job:localhost/replica:0/task:0/device:CPU:0"]]
I use python3.4, TF1.4.0
if i want use my own dataset and its size is large than 227*227?
how can i do ?
How can i properly convert one hot labels for multiple float numbers? I have tried several methods but none work.
one_hot_labels = np.zeros((batch_size, self.n_classes))
for i in range(len(labels)):
for j in range(len(labels[i])):
one_hot_labels[i][labels[j]] = 1
return images, one_hot_labels
Since i have multiple floating point numbers per image that's why i have to use another loop to iterate over its contents.
The error which i get is :
IndexError: only integers, slices (:), ellipsis (...), numpy.newaxis (None) and integer or boolean arrays are valid indices
Found this useful library but has floating point conversion errors.
Also tried manually converting labels list to numpy array and then re-evaluating, problem still exist.
I run my own dataset,it apper the fllowing error,why?\
Traceback (most recent call last):
File "finetune.py", line 109, in
horizontal_flip = True, shuffle = True)
File "/mnt/data/wxb/finetune_alexnet_with_tensorflow/datagenerator.py", line 33, in init
self.shuffle_data()
File "/mnt/data/wxb/finetune_alexnet_with_tensorflow/datagenerator.py", line 55, in shuffle_data
images = self.images.copy()
AttributeError: 'list' object has no attribute 'copy
I get 50% accuracy on the ILSVRC2012 validation set without training, which falls short of the 60% boasted in the AlexNet paper (i.e 40.7% top-1 error rate).
Any idea what could be the problem? I'm thinking this could be due to the parameters of the local_response_normalisation layer: yours seem different than the caffe AlexNet here. For instance you have alpha=1e-5 instead of 1e-4. I've been playing with those but have yet to find a configuration that gets me to 60%.
Hello,
Thanks for the share. I successfully installed and validated it with Imagenet pictures.
I also created the train.txt en val.txt files.
But when I'm trying to finetune AlexNet, I get the following error :
`WARNING:tensorflow:From`
C:/Users/Fabien/PycharmProjects/finetune_alexnet_with_tensorflow/finetune.py:98: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.
See tf.nn.softmax_cross_entropy_with_logits_v2.
2018-04-01 20:11:34.920700: I T:\src\github\tensorflow\tensorflow\core\platform\cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2018-04-01 20:11:36.529273 Start training...
2018-04-01 20:11:36.529273 Open Tensorboard at --logdir /tmp/finetune_alexnet/tensorboard
2018-04-01 20:11:36.529273 Epoch number: 1
Traceback (most recent call last):
File "C:\Program Files\JetBrains\PyCharm Community Edition 2018.1\helpers\pydev\pydevd.py", line 1664, in <module>
main()
File "C:\Program Files\JetBrains\PyCharm Community Edition 2018.1\helpers\pydev\pydevd.py", line 1658, in main
globals = debugger.run(setup['file'], None, None, is_module)
File "C:\Program Files\JetBrains\PyCharm Community Edition 2018.1\helpers\pydev\pydevd.py", line 1068, in run
pydev_imports.execfile(file, globals, locals) # execute the script
File "C:\Program Files\JetBrains\PyCharm Community Edition 2018.1\helpers\pydev\_pydev_imps\_pydev_execfile.py", line 18, in execfile
exec(compile(contents+"\n", file, 'exec'), glob, loc)
File "C:/Users/Fabien/PycharmProjects/finetune_alexnet_with_tensorflow/finetune.py", line 175, in <module>
keep_prob: dropout_rate})
File "C:\Users\Fabien\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow\python\client\session.py", line 905, in run
run_metadata_ptr)
File "C:\Users\Fabien\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow\python\client\session.py", line 1109, in _run
np_val = np.asarray(subfeed_val, dtype=subfeed_dtype)
File "C:\Users\Fabien\AppData\Local\Programs\Python\Python36\lib\site-packages\numpy\core\numeric.py", line 492, in asarray
return array(a, dtype, copy=False, order=order)
ValueError: could not convert string to float: b'C:\\Users\\Fabien\\PycharmProjects\\finetune_alexnet_with_tensorflow\\flowers_train\\class13\\image_1026.jpg'
And I can't find out where this come from and how to solve it
Thanks for helping me :)
Dear kratzert,
Thanks for the excellent work. And I am trying to fine tune with tensorflow with my own dataset. I have only images and no generated .txt files for now.
Aren't there any codes in the downloaded .zip files for generating .txt files?
hi
I found my Validation Accuracy was a constant value during 10 epochs or enven 40 epochs. I am very confused.
I found in Validate:
sess.run(validation_init_op)
.....
img_batch, label_batch = sess.run(next_batch)
but the next_batch only find in
iterator = Iterator.from_structure(tr_data.data.output_types,
tr_data.data.output_shapes)
next_batch = iterator.get_next()
so, is there any possibility a bug here, and should add like this:
iterator_val = Iterator.from_structure(val_data.data.output_types,
val_data.data.output_shapes)
next_batch_val = iterator_val.get_next()
I fine-tuned using caffe and I got 61% accuracy whereas by using this code in tensorflow I got 56% accuracy. For both cases I trained only fc8 layer. Can you explain why it is?
Hi
I met this error message when executing finetune.py. (shown as follows)
Environment: Win 7, Anaconda 4.3.0 , Python 3.6.1
Do you have any ideas about what happened?
Traceback (most recent call last):
File "C:\Anaconda3\lib\site-packages\IPython\core\interactiveshell.py", line 2898, in run_code
self.showtraceback()
File "C:\Anaconda3\lib\site-packages\IPython\core\interactiveshell.py", line 1826, in showtraceback
self._showtraceback(etype, value, stb)
File "C:\Anaconda3\lib\site-packages\ipykernel\zmqshell.py", line 554, in _showtraceback
dh.parent_header, ident=topic)
File "C:\Anaconda3\lib\site-packages\jupyter_client\session.py", line 712, in send
to_send = self.serialize(msg, ident)
File "C:\Anaconda3\lib\site-packages\jupyter_client\session.py", line 607, in serialize
content = self.pack(content)
File "C:\Anaconda3\lib\site-packages\jupyter_client\session.py", line 103, in
ensure_ascii=False, allow_nan=False,
File "C:\Anaconda3\lib\site-packages\zmq\utils\jsonapi.py", line 43, in dumps
s = s.encode('utf8')
UnicodeEncodeError: 'utf-8' codec can't encode character '\udcc0' in position 2120: surrogates not allowed
It seems that
tf.nn.softmax_cross_entropy_with_logits
needs outputs to be a softmax,right?
I'm just curious about that,no offence
I get this exception at the trainning part, but it doesn't happen always (it looks a random issue):
2018-04-18 15:02:15.902533 Start validation Traceback (most recent call last): File "finetune.py", line 188, in <module> print("{} Start validation".format(datetime.now())) OSError: raw write() returned invalid length 90 (should have been between 0 and 45)
I'm using Windows 10 64 bits with Python 3.5.2 :: Anaconda 4.2.0 (64-bit) and Tensorflow 1.2.1.
My dataset is Dogs VS Cats.
Besides, I'd like to recover the already trained model, since it's very upsetting getting this crash after some epoches. As you indicated, I tried to change the load_initial_weights line to this line:
saver.restore(sess, "/path/to/checkpoint/model.ckpt")
But I have three cktp files (index, data and meta). What should I do?
Thanks!
img_bgr = img_centered[:, :, ::-1]
Hi bro, I am not familiar with python. Could you explain some about slice operation of list, as above from your code.
Thanks!
Hi,
For a two-class problem, I got the following result.
The cross-entropy is very very small, but I can not get high accuracy.
I traced the code of computing loss value (following code),
"loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=score, labels=y))"
and found that the score need to be " Unscaled log probabilities" (https://www.tensorflow.org/api_docs/python/tf/nn/softmax_cross_entropy_with_logits)
So it seems that you need to perform a log function to "score" first before feeding it to softmax function?
Please correct me if I'm wrong, and many thanks.
I have successfully trained alexnet for dogs vs cats dataset but when I run tensorboard I can only visualize this page:
The rest of the pages are inactive:
I have tensorflow 1.3.0, and python 3.6.3 and I am using windows 10. Do you have any idea what can cause the problem?
I get the shape like this:
conv1
0 conv1 (11, 11, 3, 96)
0 conv1 (96,)
conv2
1 conv2 (5, 5, 48, 256)
1 conv2 (256,)
conv3
2 conv3 (3, 3, 256, 384)
2 conv3 (384,)
conv4
3 conv4 (3, 3, 192, 384)
3 conv4 (384,)
conv5
4 conv5 (3, 3, 192, 256)
4 conv5 (256,)
how to convert shape conv2(5,5,48,256) to (5,5,96,256)
conv4 (3,3,192,384) to (3,3,384,384)
conv5(3,3,192,256) to (3,3,384,256)
I follow your step to run the code but ended up with different result,the logs file can be save correctly but the scalar and histogram data can't be shown on the tensorboard,while we can see the graph with tensorboard,please help to solve the problem ,thanks in advance!
I want to run file validate_alexnet_on_imagenet.ipynb to test the implement of AlexNet. When I run it on my Windows PC, it works well. But when I run it on my Linux server, which is much more powerful than my PC, it's stuck for a while and output a ResourceExhaustedError, which I believe occurred in this line:
sess.run(tf.global_variables_initializer())
Full output is shown below:
---------------------------------------------------------------------------
ResourceExhaustedError Traceback (most recent call last)
/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py in _do_call(self, fn, *args)
1322 try:
-> 1323 return fn(*args)
1324 except errors.OpError as e:
/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py in _run_fn(session, feed_dict, fetch_list, target_list, options, run_metadata)
1301 feed_dict, fetch_list, target_list,
-> 1302 status, run_metadata)
1303
/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/errors_impl.py in __exit__(self, type_arg, value_arg, traceback_arg)
472 compat.as_text(c_api.TF_Message(self.status.status)),
--> 473 c_api.TF_GetCode(self.status.status))
474 # Delete the underlying status object from memory otherwise it stays alive
ResourceExhaustedError: OOM when allocating tensor with shape[4096]
[[Node: fc6/biases/Initializer/random_uniform/RandomUniform = RandomUniform[T=DT_INT32, _class=["loc:@fc6/biases"], dtype=DT_FLOAT, seed=0, seed2=0, _device="/job:localhost/replica:0/task:0/device:GPU:0"](fc6/biases/Initializer/random_uniform/shape)]]
During handling of the above exception, another exception occurred:
ResourceExhaustedError Traceback (most recent call last)
<ipython-input-4-eb8e619c57ae> in <module>()
2
3 # Initialize all variables
----> 4 sess.run(tf.global_variables_initializer())
5
6 # Load the pretrained weights into the model
/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py in run(self, fetches, feed_dict, options, run_metadata)
887 try:
888 result = self._run(None, fetches, feed_dict, options_ptr,
--> 889 run_metadata_ptr)
890 if run_metadata:
891 proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)
/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py in _run(self, handle, fetches, feed_dict, options, run_metadata)
1118 if final_fetches or final_targets or (handle and feed_dict_tensor):
1119 results = self._do_run(handle, final_targets, final_fetches,
-> 1120 feed_dict_tensor, options, run_metadata)
1121 else:
1122 results = []
/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py in _do_run(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)
1315 if handle is None:
1316 return self._do_call(_run_fn, self._session, feeds, fetches, targets,
-> 1317 options, run_metadata)
1318 else:
1319 return self._do_call(_prun_fn, self._session, handle, feeds, fetches)
/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/client/session.py in _do_call(self, fn, *args)
1334 except KeyError:
1335 pass
-> 1336 raise type(e)(node_def, op, message)
1337
1338 def _extend_graph(self):
ResourceExhaustedError: OOM when allocating tensor with shape[4096]
[[Node: fc6/biases/Initializer/random_uniform/RandomUniform = RandomUniform[T=DT_INT32, _class=["loc:@fc6/biases"], dtype=DT_FLOAT, seed=0, seed2=0, _device="/job:localhost/replica:0/task:0/device:GPU:0"](fc6/biases/Initializer/random_uniform/shape)]]
Caused by op 'fc6/biases/Initializer/random_uniform/RandomUniform', defined at:
File "/usr/lib/python3.5/runpy.py", line 193, in _run_module_as_main
"__main__", mod_spec)
File "/usr/lib/python3.5/runpy.py", line 85, in _run_code
exec(code, run_globals)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/ipykernel_launcher.py", line 16, in <module>
app.launch_new_instance()
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/traitlets/config/application.py", line 658, in launch_instance
app.start()
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/ipykernel/kernelapp.py", line 477, in start
ioloop.IOLoop.instance().start()
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/zmq/eventloop/ioloop.py", line 177, in start
super(ZMQIOLoop, self).start()
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tornado/ioloop.py", line 888, in start
handler_func(fd_obj, events)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tornado/stack_context.py", line 277, in null_wrapper
return fn(*args, **kwargs)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/zmq/eventloop/zmqstream.py", line 440, in _handle_events
self._handle_recv()
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/zmq/eventloop/zmqstream.py", line 472, in _handle_recv
self._run_callback(callback, msg)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/zmq/eventloop/zmqstream.py", line 414, in _run_callback
callback(*args, **kwargs)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tornado/stack_context.py", line 277, in null_wrapper
return fn(*args, **kwargs)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/ipykernel/kernelbase.py", line 283, in dispatcher
return self.dispatch_shell(stream, msg)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/ipykernel/kernelbase.py", line 235, in dispatch_shell
handler(stream, idents, msg)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/ipykernel/kernelbase.py", line 399, in execute_request
user_expressions, allow_stdin)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/ipykernel/ipkernel.py", line 196, in do_execute
res = shell.run_cell(code, store_history=store_history, silent=silent)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/ipykernel/zmqshell.py", line 533, in run_cell
return super(ZMQInteractiveShell, self).run_cell(*args, **kwargs)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/IPython/core/interactiveshell.py", line 2728, in run_cell
interactivity=interactivity, compiler=compiler, result=result)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/IPython/core/interactiveshell.py", line 2850, in run_ast_nodes
if self.run_code(code, result):
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/IPython/core/interactiveshell.py", line 2910, in run_code
exec(code_obj, self.user_global_ns, self.user_ns)
File "<ipython-input-3-5d95ec9c311f>", line 9, in <module>
model = AlexNet(x, keep_prob, 1000, [])
File "/home1/grayson/finetune_alexnet_with_tensorflow/alexnet.py", line 56, in __init__
self.create()
File "/home1/grayson/finetune_alexnet_with_tensorflow/alexnet.py", line 82, in create
fc6 = fc(flattened, 6*6*256, 4096, name='fc6')
File "/home1/grayson/finetune_alexnet_with_tensorflow/alexnet.py", line 177, in fc
biases = tf.get_variable('biases', [num_out], trainable=True)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variable_scope.py", line 1203, in get_variable
constraint=constraint)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variable_scope.py", line 1092, in get_variable
constraint=constraint)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variable_scope.py", line 425, in get_variable
constraint=constraint)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variable_scope.py", line 394, in _true_getter
use_resource=use_resource, constraint=constraint)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variable_scope.py", line 805, in _get_single_variable
constraint=constraint)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variables.py", line 213, in __init__
constraint=constraint)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variables.py", line 303, in _init_from_args
initial_value(), name="initial_value", dtype=dtype)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/variable_scope.py", line 779, in <lambda>
shape.as_list(), dtype=dtype, partition_info=partition_info)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/init_ops.py", line 445, in __call__
shape, -limit, limit, dtype, seed=self.seed)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/random_ops.py", line 240, in random_uniform
shape, dtype, seed=seed1, seed2=seed2)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/ops/gen_random_ops.py", line 473, in _random_uniform
name=name)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/op_def_library.py", line 787, in _apply_op_helper
op_def=op_def)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 2956, in create_op
op_def=op_def)
File "/home1/grayson/tensorflow/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 1470, in __init__
self._traceback = self._graph._extract_stack() # pylint: disable=protected-access
ResourceExhaustedError (see above for traceback): OOM when allocating tensor with shape[4096]
[[Node: fc6/biases/Initializer/random_uniform/RandomUniform = RandomUniform[T=DT_INT32, _class=["loc:@fc6/biases"], dtype=DT_FLOAT, seed=0, seed2=0, _device="/job:localhost/replica:0/task:0/device:GPU:0"](fc6/biases/Initializer/random_uniform/shape)]]
Anyone has an idea about it?
Hi
I have fine tune my model using this code, now i can predeict new data through trained model.
but i need to extract features of full connected layer fc7.
how it can be possible
Hi,
I finished training Alexnet and output some .meta files in checkpoints folder. Then I want to predicted a picture with the new model parameters. I followed the validate_alexnet_on_imagenet.ipynb file step by step and changed "model = AlexNet(x, keep_prob, 1000, [])" to "model = AlexNet(x, keep_prob, 2, [])", as in my processes the num_classes are 2. The error message is as follows:
Dimension 1 in both shapes must be equal, but are 2 and 1000. Shapes are [4096,2] and [4096,1000]. for 'fc8_1/Assign' (op: 'Assign') with input shapes: [4096,2], [4096,1000].
I am not sure whether I need to change any other codes in validate_alexnet_on_imagenet.ipynb
Thank you
I run the finetune.py
I get a error AttributeError: 'list' object has no attribute 'copy'
`File "/usr/local/lib/python2.7/dist-packages/spyder/utils/site/sitecustomize.py", line 880, in runfile
execfile(filename, namespace)
File "/usr/local/lib/python2.7/dist-packages/spyder/utils/site/sitecustomize.py", line 94, in execfile
builtins.execfile(filename, *where)
File "/home/lan/tensorflow/finetune_alexnet_with_tensorflow-5d751d62eb4d7149f4e3fd465febf8f07d4cea9d/finetune.py", line 109, in
horizontal_flip = True, shuffle = True)
File "datagenerator.py", line 31, in init
self.shuffle_data()
File "datagenerator.py", line 53, in shuffle_data
images = self.images.copy()
AttributeError: 'list' object has no attribute 'copy'`
I am trying to implement a siamese network for the first time. I don't have an experience with variable sharing. I don't know why I become this Error. Any help will be appreciated
`from future import division, print_function, absolute_import
import tensorflow as tf
import tflearn
from tflearn.layers.core import input_data, dropout, fully_connected
from tflearn.layers.conv import conv_2d, max_pool_2d
from tflearn.layers.normalization import local_response_normalization
from tflearn.layers.estimator import regression
import tflearn.datasets.mnist as mnist
X, Y, testX, testY = mnist.load_data(one_hot=True)
X = X.reshape([-1, 28, 28, 1])
testX = testX.reshape([-1, 28, 28, 1])
def tower_network(reuse= False):
network = tflearn.input_data(shape=(None,28,28,1))
network = tflearn.conv_2d(network, 32,1, activation='relu',reuse=reuse, scope='conv1')
network = tflearn.conv_2d(network, 64,1, activation='relu',reuse=reuse, scope='conv2')
network = tflearn.conv_2d(network, 128,1, activation='relu',reuse=reuse, scope='conv3')
network = tflearn.max_pool_2d(network, 2, strides=2)
network = tflearn.fully_connected(network, 512, activation='relu',reuse=reuse, scope='fc1')
network = tflearn.dropout(network, 0.5)
return network
def similarity_network( net1, net2):
num_classes = 2
network = tflearn.merge([net1,net2], mode='concat', axis=1, name='Merge') # merge net1 and net2 networks
# fully connected layers
network = tflearn.fully_connected(network, 2048, activation='relu')
network = tflearn.dropout(network, 0.5)
network = tflearn.fully_connected(network, 2048, activation='relu')
network = tflearn.dropout(network, 0.5)
# softmax layers
network = tflearn.fully_connected(network, num_classes, activation='softmax')
return network
net1 = tower_network()
net2 = tower_network(reuse=True)
#similarity network
network = similarity_network( net1, net2)
#output layer
#network = tflearn.regression(network, optimizer='sgd', loss='hinge_loss', learning_rate=0.02)
network = tflearn.regression(network, optimizer='sgd', loss='categorical_crossentropy', learning_rate=0.02)
model = tflearn.DNN(network, tensorboard_verbose=0)
model.fit({'input': X}, {'target': Y}, n_epoch=20,
validation_set=({'input': testX}, {'target': testY}),
snapshot_step=100, show_metric=True, run_id='convnet_mnist')`
Hello there
In the datagenerator I find the VGG ,can you tell we what it is .
And I find that you use it to make the data centerned
But for the variance is required to equal to one , how to solve it
I really do not know how to do the preprocess , may I was wrong
Hope for response , thank you!
Dear kratzert,
First, thanks for your amazing work !
I would like to use your code to solve a regression problem, do have some time to help me ?
Best.
Hi, this is not an issue but a question:
watching your code i wanna ask you why you pass padding = VALID when you call function here
and in your definition of that function you pass padding = SAME
Same thing about conv
Thanks for the huge work.
Hope you can solve my doubt
hi, I did some math computation and check the original paper:
the first convolution filter is [11,11,3,96], there is no maxpool,
second convolution filter is [5,5,96,256], and then we get maxpool
but in your code
conv1 = conv(self.X, 11, 11, 96, 4, 4, padding='VALID', name='conv1')
norm1 = lrn(conv1, 2, 2e-05, 0.75, name='norm1')
pool1 = max_pool(norm1, 3, 3, 2, 2, padding='VALID', name='pool1')
# 2nd Layer: Conv (w ReLu) -> Lrn -> Pool with 2 groups
conv2 = conv(pool1, 5, 5, 256, 1, 1, groups=2, name='conv2')
norm2 = lrn(conv2, 2, 2e-05, 0.75, name='norm2')
pool2 = max_pool(norm2, 3, 3, 2, 2, padding='VALID', name='pool2')
, I didn't get it, which is not the same as the picture, can you explain?
I cannot obtain the test accuracy from tensor-board. Can you provide any example of Embedding Visualization, please ?
I think you talk about GPU on your blog
hi,
I want to add dataaugmentation in datagenerator.py. According to your comments in def _parse_function_train(self, filename, label):
"""
Dataaugmentation comes here.
"""
say I want to flip_up_and down and flip_left_right, in this way, and corp from 4 corner, one input image could have 6 output images.
so my question is: does the _parse_function_train() function still work for data.map(self._parse_function_train,....). if not, how I add my dataaugmentation in the code?
thanks!
My point is in _parse_function_train() function, one input image could have 6 output images. Is that ok?
I get the following error from datagenerator.py:
error: (-215) ssize.width > 0 && ssize.height > 0 in function resize
finetune_alexnet_with_tensorflow-master/datagenerator.py", line 150, in next_batch img = cv2.resize(img, (self.scale_size[0], self.scale_size[0]))
I debugged code as follows:
img = cv2.imread(paths[i])
print(img);
if img is None:
print (paths[i] + " : fail to read")
else:
print("Image is read");
exit();
Error suggests that image is not read while in a separate program i verified that image is read:
import numpy as np
import cv2
img = cv2.imread('images/ld2.png')
cv2.imwrite('messigray.png',img)
Followed this solution but couldn't work.
Hi! Thanks for your wonderful program!
In the "self._parse_function_train" function ,do you load all of your training dataset ?
I want to apply your code to video dataset，and should I load all of the training dataset in this function?
thanks!
Thanks for your contribution!
when i reuse your code, and i want to train fc6, fc7,fc8 , but the result:
Cross entropy = nan
if only train fc7,fc8 , it works well.
also i have test earlier layers ,, like conv1,conv2.., it still appear nan
do have have any suggestion to help me fix this?
Hi, I am having some trouble running the third cell of the notebook. Error is:
ValueErrorTraceback (most recent call last)
<ipython-input-7-f7a1b7dd0c14> in <module>()
7
8 #create model with default config ( == no skip_layer and 1000 units in the last layer)
----> 9 model = AlexNet(x, keep_prob, 1000, [])
10
11 #define activation of last layer as score
/mnt/ilcompf6d0/user/txiao/DockerFiles/finetune_alexnet_with_tensorflow/alexnet.py in __init__(self, x, keep_prob, num_classes, skip_layer, weights_path)
39
40 # Call the create function to build the computational graph of AlexNet
---> 41 self.create()
42
43 def create(self):
/mnt/ilcompf6d0/user/txiao/DockerFiles/finetune_alexnet_with_tensorflow/alexnet.py in create(self)
44
45 # 1st Layer: Conv (w ReLu) -> Pool -> Lrn
---> 46 conv1 = conv(self.X, 11, 11, 96, 4, 4, padding = 'VALID', name = 'conv1')
47 pool1 = max_pool(conv1, 3, 3, 2, 2, padding = 'VALID', name = 'pool1')
48 norm1 = lrn(pool1, 2, 2e-05, 0.75, name = 'norm1')
/mnt/ilcompf6d0/user/txiao/DockerFiles/finetune_alexnet_with_tensorflow/alexnet.py in conv(x, filter_height, filter_width, num_filters, stride_y, stride_x, name, padding, groups)
131 with tf.variable_scope(name) as scope:
132 # Create tf variables for the weights and biases of the conv layer
--> 133 weights = tf.get_variable('weights', shape = [filter_height, filter_width, input_channels/groups, num_filters])
134 biases = tf.get_variable('biases', shape = [num_filters])
135
/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.pyc in get_variable(name, shape, dtype, initializer, regularizer, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter)
1047 collections=collections, caching_device=caching_device,
1048 partitioner=partitioner, validate_shape=validate_shape,
-> 1049 use_resource=use_resource, custom_getter=custom_getter)
1050 get_variable_or_local_docstring = (
1051 """%s
/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.pyc in get_variable(self, var_store, name, shape, dtype, initializer, regularizer, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter)
946 collections=collections, caching_device=caching_device,
947 partitioner=partitioner, validate_shape=validate_shape,
--> 948 use_resource=use_resource, custom_getter=custom_getter)
949
950 def _get_partitioned_variable(self,
/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.pyc in get_variable(self, name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter)
354 reuse=reuse, trainable=trainable, collections=collections,
355 caching_device=caching_device, partitioner=partitioner,
--> 356 validate_shape=validate_shape, use_resource=use_resource)
357
358 def _get_partitioned_variable(
/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.pyc in _true_getter(name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource)
339 trainable=trainable, collections=collections,
340 caching_device=caching_device, validate_shape=validate_shape,
--> 341 use_resource=use_resource)
342
343 if custom_getter is not None:
/usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/variable_scope.pyc in _get_single_variable(self, name, shape, dtype, initializer, regularizer, partition_info, reuse, trainable, collections, caching_device, validate_shape, use_resource)
651 " Did you mean to set reuse=True in VarScope? "
652 "Originally defined at:\n\n%s" % (
--> 653 name, "".join(traceback.format_list(tb))))
654 found_var = self._vars[name]
655 if not shape.is_compatible_with(found_var.get_shape()):
ValueError: Variable conv1/weights already exists, disallowed. Did you mean to set reuse=True in VarScope? Originally defined at:
File "alexnet.py", line 133, in conv
weights = tf.get_variable('weights', shape = [filter_height, filter_width, input_channels/groups, num_filters])
File "alexnet.py", line 46, in create
conv1 = conv(self.X, 11, 11, 96, 4, 4, padding = 'VALID', name = 'conv1')
File "alexnet.py", line 41, in __init__
self.create()
This error was in both the May 22 commit (TF 1.0) and the latest commit, TF 1.12rc0. I am on TF 1.1.0.
Cheers!
Can you suggest a code snippet for multiple columns in the training file. Example train.txt contains one int column, what if i have more than one floating columns?
Thanks
Hello，recently I used your code to test image classifying.
Now,there are some stand errors.You can see below:
As for my train file, it is like this:
This task is important for me, but I am still puzzled in it.
Could you help solve this problem? Thanks to you in advance.
Hello,
Can you tell me the reason why you are converting RGB images to BGR in datagenerator.py? if I keep images as RGB what will happen?
img_bgr = img_centered[:, :, ::-1]
Hi, the Alexnet paper listed PCA as a data augmentation method, I'm wondering whether this is implemented in your model as I can't seem to find it.
Hey, so I finished fine tuning my model, but I haven't really used tensorflow that much before. I understand how to load in the images I want to test on, but how do I use the checkpoint file to classify them?
Also thoughts on parameters for this dataset; 330 training images of Benign and Malignant, along with a validation and testing set of 40 images. I changed learning rate to 0.005 as well as batch size to 40, any other thoughts? I'm still training the last two layers, I'm not sure whether this is the right move.
Thanks!!
I got a lot of information to implement alexnet, thank you.
But two implementations are not included in your blog and code but included in alexnet.
Is it intended?
Hello,Kratzert. Your blog and code are very nice,I am new to tensorflow and I learned a lot from them.
Did you try the class: tf.estimator?I found that you defined modes for training and inference in filedatagenerator.py, I refered to tensorflow tutorials and found tf.estimator can implement the same function.Why do you prefer not using this class?Maybe it has some disadvantages for summary operations?
: )
I run fintune.py,but I get some error.Could someone plz tell me what's the problem and how to modify the code to solve it? Thank you
the error is flowing:
tensorflow.python.framework.errors_impl.InvalidArgumentError: Nan in summary histogram for: fc8/biases_0
hello , I meet trouble, in the train/val set import .
I want to know how to op, I can't overcome the problem.
I am design my deadline project, so I want to how op the Net.
I'm studying deep learning and as I learn it, dropout rate when less than 1, say 0.5 for example, then during testing it's traditional to keep all the units and multiply the outputs by 0.5 in order to make a good inference. However, I'm not sure if it's my unawareness or just a bug, it seems that the mechanism has not been implemented, I hope you can check it.
Hello,
Thanks for the code. Actually, This is not an issue, It's just a question :)
I want to know how the classification model is working which, I need confusion matrix in order to understand. Tensorflow has a function for it which is (tf.confusion_matrix(labels,predictions,num_classes=None) but I am confused about what should be the labels and predictions values?
Thanks.
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.