NotFoundError (see above for traceback): ./JPEGImages/2007_000033.jpg

[jjcao]

I can perform inference over my own images via python inference.py ./test.jpg ./pretrained/deeplab_resnet.ckpt

But when I evaluate the pre-trained model: python evaluate.py --data-dir . --data-list ./dataset/val.txt --restore-from pretrained/deeplab_resnet.ckpt

I met the error:

Restored model parameters from pretrained/deeplab_resnet.ckpt 2017-08-06 20:55:44.006616: W tensorflow/core/framework/op_kernel.cc:1158] Not found: ./SegmentationClassAug/2007000033.png Traceback (most recent call last): File "evaluate.py", line 128, in main() File "evaluate.py", line 120, in main preds, = sess.run([pred, update_op]) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 789, in run run_metadata_ptr) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 997, in _run feed_dict_string, options, run_metadata) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1132, in _do_run target_list, options, run_metadata) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1152, in _do_call raise type(e)(node_def, op, message) tensorflow.python.framework.errors_impl.NotFoundError: ./SegmentationClassAug/2007_000033.png [[Node: create_inputs/ReadFile_1 = ReadFile_device="/job:localhost/replica:0/task:0/cpu:0"]]

Caused by op u'create_inputs/ReadFile_1', defined at: File "evaluate.py", line 128, in main() File "evaluate.py", line 79, in main coord) File "/Users/jjcao/Downloads/deeplabV2/deeplab_resnet/image_reader.py", line 168, in init self.image, self.label = read_images_from_disk(self.queue, self.input_size, random_scale, random_mirror, ignore_label, img_mean) File "/Users/jjcao/Downloads/deeplabV2/deeplab_resnet/image_reader.py", line 112, in read_images_from_disk img_contents = tf.read_file(input_queue[0]) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/ops/gen_io_ops.py", line 200, in read_file result = _op_def_lib.apply_op("ReadFile", filename=filename, name=name) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 767, in apply_op op_def=op_def) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2506, in create_op original_op=self._default_original_op, op_def=op_def) File "/Users/jjcao/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1269, in init self._traceback = _extract_stack()

NotFoundError (see above for traceback): ./SegmentationClassAug/2007_000033.png [[Node: create_inputs/ReadFile_1 = ReadFile_device="/job:localhost/replica:0/task:0/cpu:0"]]

======================= But ./SegmentationClassAug/2007_000033.png is there. Why image_reader.py fails to read it?

Thank you.

[jjcao]

I wrote a simple test.py to read './SegmentationClassAug/2007_000033.png'. It works. Why tf.read_file failed when it is called by evaluate.py?

import tensorflow as tf import numpy as np

def main(): sess = tf.InteractiveSession() pngFILE = './SegmentationClassAug/2007_000033.png' label_contents = tf.read_file(pngFILE) label = tf.image.decode_png(label_contents, channels=1) sess.close()

if name == 'main': main()

[YCJGG]

The "\n" is different in windows and linux. Just try your code in Linux

[DrSleep]

Not sure about this one, as this img_contents = tf.read_file(input_queue[0]) and this ./SegmentationClassAug/2007_000033.png confuses me: input_queue[0] should point to the JPG file. Something is wrong when reading in your files.

First, make sure that the lists are processed correctly: for example, print the outputs of this function for debugging: https://github.com/DrSleep/tensorflow-deeplab-resnet/blob/master/deeplab_resnet/image_reader.py#L91