OutOfRangeError (see above for traceback):

[sed0724963]

what it is mean??object_detection module is equal a file of "object_detection" ?or not?

[VinceKumar]

You have to rerun instructions from Protobuf Compilation, Add Libraries to PYTHONPATH everytime you close the terminal.

link

[sed0724963]

I was written at PATH.. 。 it's your mean??

[k-w-w]

pwd should be replaced with the path to the models/research folder.

[VinceKumar]

shouldn't it also be in your .bashrc file?

[sed0724963]

Now, it's my new question,I try to renew the command line ,it's become this ..:

No such file or directory

[k-w-w]

You're missing a space between --logtostderr and --pipeline_config_path.

[sed0724963]

I renew the command line , Take away the $ mark become:python object_detection/train.py --logtostderr --pipeline_config_path=object_detection/data/faster_rcnn_resnet101_pets.config --train_dir=object_detection/data/
and problem become this...

[VinceKumar]

#3705

[sed0724963]

a new problem...

[k-w-w]

That appears to be a bug in tf.contrib.slim. Are you able to upgrade to Tensorflow 1.7 or later?

[sed0724963]

OK, I'll trying now

[sed0724963]

it's successful?? so faster?

[k-w-w]

It looks like it didn't train at all. Did you prepare the data as instructed by the README?

[sed0724963]

I think I have...

[sed0724963]

If you ' ll allow a slight digression, how did you know about my tensorflow version is too low?? the previous question

[k-w-w]

I'm not too familiar with the object detection model, so I'm unsure if that's the expected train directory structure. @derekjchow , can you help with this? Thanks!

@sed0724963 I saw the tf.contrib.slim bug on the r1.6 branch in the source code. The bug was fixed in r1.7. (I also probably could have used the release notes to figure out when the bug was fixed)

[sed0724963]

but I only have image of my terminal , how did you know ? yourself seen the souce code at before? and how to know my question is same??

[k-w-w]

The error local variable 'total_loss' referenced before assignment originated from the file that I linked (learning.py), so I look at the code history and found that there was a bug in r1.6 and before.

[sed0724963]

Unsuccessful TensorSliceReader constructor??
I only have model.ckpt ,but it tell me must model.ckpt-0?? why??

[SpiralBeing]

@sed0724963 The same issue..You solved it?

[sed0724963]

@FreedoomFighter what's issue ? sorry I forgot....,but now I can work! so I think I was solved.