yeephycho / tensorflow_input_image_by_tfrecord Goto Github PK

Hi,

I'm trying to run read_tfrecord_data.py after running build_image_data.py successfully. read_tfrecord_data.py giving me an error message as follow:
File "read_tfrecord_data.py", line 27
self.image = tf.Variable([], dtype = tf.string)
^
TabError: inconsistent use of tabs and spaces in indentation

I didn't write the code nor changed it. I used clone to get the files.

Is there any ideas on how to solve this issue.

Best,
IA

博主，你好，在src/dog文件夹下，有一个test文件，我怀着对未知事物充满好奇的心情，用激动地颤抖的手，握着鼠标，虔诚地点开了它。果然，没让我失望，屏幕赫赫出现一句：“知识就是力量”。于是，我忍不住开了这个issue。

Hi I'm trying to get your program to run on the sample data and I keep getting an error. I've tried using python 3.6 with tf 1.4 and even python 3.5 with tf 0.12 but I get an error. I'm on MacOS but I can VM if I need to.

Hello, I learned a lot from your this example. I tried to modefied the code to read the tfrecord file and feed into the graph, but I met some problems.
Can you help solve my problem?

https://stackoverflow.com/questions/46985411/tensorflow-read-tfrecord-without-a-graph

I run the examples in README
"python build_image_data.py"
OK
Wrote 150 images to ./train-00000-of-00002
Wrote 150 images to ./train-00001-of-00002

Then I run next command

"python read_tfrecord_data.py --image_number=300 --class_number=3 --image_height=299 --image_width=299"
20 files were found under current folder.
Please be noted that only files end with '*.tfrecord' will be load!
Cannot find any tfrecord files, please check the path.

How does the creation of filename.tfrecord happen?

python build_image_data.py
Saving results to ./
Determining list of input files and labels from ./.
Found 302 JPEG files across 3 labels inside ./.
Launching 2 threads for spacings: [[0, 151], [151, 302]]
2018-01-23 00:37:19.690943 [thread 0]: Wrote 0 images to 151 shards.
2018-01-23 00:37:19.691943 [thread 1]: Wrote 0 images to 151 shards.
2018-01-23 00:37:20.692609: Finished writing all 302 images in data set.
Determining list of input files and labels from ./.
Found 302 JPEG files across 3 labels inside ./.
Launching 2 threads for spacings: [[0, 151], [151, 302]]
2018-01-23 00:37:22.013490: Finished writing all 302 images in data set.
Exception in thread Thread-3:
Traceback (most recent call last):
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\threading.py", line 914, in _bootstrap_inner
self.run()
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\threading.py", line 862, in run
self._target(*self._args, **self._kwargs)
File "build_image_data.py", line 264, in _process_image_files_batch
image_buffer, height, width = _process_image(filename, coder)
File "build_image_data.py", line 203, in _process_image
image_data = f.read()
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 125, in read
pywrap_tensorflow.ReadFromStream(self._read_buf, length, status))
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 93, in _prepare_value
return compat.as_str_any(val)
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\util\compat.py", line 106, in as_str_any
return as_str(value)
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\util\compat.py", line 84, in as_text
return bytes_or_text.decode(encoding)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte

Exception in thread Thread-4:
Traceback (most recent call last):
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\threading.py", line 914, in _bootstrap_inner
self.run()
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\threading.py", line 862, in run
self._target(*self._args, **self._kwargs)
File "build_image_data.py", line 264, in _process_image_files_batch
image_buffer, height, width = _process_image(filename, coder)
File "build_image_data.py", line 203, in _process_image
image_data = f.read()
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 125, in read
pywrap_tensorflow.ReadFromStream(self._read_buf, length, status))
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\lib\io\file_io.py", line 93, in _prepare_value
return compat.as_str_any(val)
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\util\compat.py", line 106, in as_str_any
return as_str(value)
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\util\compat.py", line 84, in as_text
return bytes_or_text.decode(encoding)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 3131: invalid start byte

2018-01-23 00:37:19.430514: W C:\tf_jenkins\home\workspace\rel-win\M\windows\PY\35\tensorflow\core\platform\cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.

python read_tfrecord_data.py --image_number=300 --class_number=3 --image_height=299 --image_width=299
20 files were found under current folder.
Please be noted that only files end with '*.tfrecord' will be load!
Cannot find any tfrecord files, please check the path.
Traceback (most recent call last):
File "read_tfrecord_data.py", line 70, in
shuffle = True)
File "C:\Users\Francis\AppData\Local\Programs\Python\Python35\lib\site-packages\tensorflow\python\training\input.py", line 217, in string_input_producer
raise ValueError(not_null_err)
ValueError: string_input_producer requires a non-null input tensor

I am trying to use your tutorial.
So in the build_image_data.py file we have the following:

tf.app.flags.DEFINE_string('train_directory', './',
                           'Training data directory')
tf.app.flags.DEFINE_string('validation_directory', './',
                           'Validation data directory')
tf.app.flags.DEFINE_string('output_directory', './',
                           'Output data directory')

Should I do anything after I do "cd ../tensorflow_input_image_by_tfrecord/src"? Do I need to create folders namely Output data directory, Validation data directory, Training data directory?

When I just run the code as it is explained, I receive the error:
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x80 in position 110: invalid start byte

Thanks for the tutorial!
It is really useful.
I am using Tensorflow on windows via Python 3.6 with Anaconda.
When I run build_image_data I received the following error:
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte

Any idea what I need to do?

Thanks!

How exactly do I create a labels file? I know what should go in it (valid labels, line by line), but what do I need to name it and where do I store it? At this point I have 6 folders -- "Bullfrogs," "Salamanders," "Goldfish," "Bullfrogs_t" (testing images of bullfrogs), "Salamanders_t" and "Goldfish_t". I want to create a classification algorithm in TensorFlow (don't actually need it, just practicing for when I have my real data). Can you help me? Thanks so much for creating this script.

Hi, @yeephycho
After run build_image_data.py，output two file:

1. train-00000-of-00002
2. train-00001-of-00002
   But in read_tfrecord_data.py, the file should be .tfrecord format, ie, when run:
   python read_tfrecord_data.py --image_number=300 --class_number=3 --image_height=299 --image_width=299

generate an error:
Please be noted that only files end with '*.tfrecord' will be load!
Cannot find any tfrecord files, please check the path.
Because it cannot find any *.tfrecord file.
Solution:
Modify the line 253 of build_image_data.py :
output_filename = '%s-%.5d-of-%.5d' % (name, shard, num_shards)
to
output_filename = '%s-%.5d-of-%.5d.tfrecord' % (name, shard, num_shards)

Need the code to predict an image for the model trained using flower_train_cnn.py. Was trying to use conv_mnist_inference.py and match the parameters in the evaluation logic in flower_train_cnn.py.
BATCH_SIZE = 10
x = tf.placeholder(tf.float32, shape=[BATCH_SIZE, 5])
x_image = tf.reshape(x, [-1, IMAGE_SIZE, IMAGE_SIZE, 3])

Getting the following error
ValueError: Dimension size must be evenly divisible by 150528 but is 50 for 'Reshape' (op: 'Reshape') with input shapes: [10,5], [4] and with input tensors computed as partial shapes: input[1] = [?,224,224,3].

Also i suppose the same flower_train_cnn.py would work for training any tfrecord files we create using the images and label,txt by running the command build_image_data.py.
Was a bit confused http://yeephycho.github.io/2016/08/15/image-data-in-tensorflow/ following the link. and hence need your guidance.

Thanking you.