vonclites / squeezenet Goto Github PK

deleted

I found there is a details in the paper, that is "• ReLU (Nair & Hinton, 2010) is applied to activations from squeeze and expand layers." But I can't find any use of ReLU in squeezenet.py. Are there some reason for without ReLU?

Hi, Thanks for your excellent work! I tried to implement it on a raspberry pi for a kitchen system to detect fruits and vegetables. Thus I need to retrain the network for around 20 classes. I already prepared the dataset from ImageNet but I am not sure how can I feed them to your scripts?

Hello @vonclites,

I want to know if this script needs only image size of 224x224 while creating a tfrecord file?
Or will it work for any size image?
Please, let me know

Thanks

@vonclites Hello,
Thanks for the architecture in tensorflow. I have trained squeezenet on caffe with my custom dataset and being very new to tensorflow(started it 2 days back) I am facing difficulty in understanding how to use this architecture for training on my custom data. I have trained it with Cifar but now I want to train it with my on custom data.
Any directions on how to train it understanding my non-familarity with the platform might be very helpful.

Thanks!

Hi Vonclites, thanks for sharing the squeezenet implementation. I was playing around with it and I noticed that the output to the inference on eval is dependent on the batch size of the input. For instance, with a batch size of 1 the output it different than batch size of 5. Looking into it, but perhaps you may have seen this.

Squeezenet Training Program: error: the following arguments are required: --model_dir, --train_tfrecord_filepaths, --validation_tfrecord_filepaths, --network, --num_classes, --num_gpus, --batch_size, --num_input_threads, --shuffle_buffer

Hi, does anybody know how to set the arguments? I did not find relating files in the repo, for example the tfrecord_filepaths...

I am not able to import slim. What is this and where did you find this library?

It always throws the following exception,can you give some tips to fix this issue?
I just tried it on tensorflow 1.9. I am not sure whether it works on other version.

Traceback (most recent call last):
File "train_squeezenet.py", line 184, in
run()
File "train_squeezenet.py", line 180, in run
_run(args)
File "train_squeezenet.py", line 107, in _run
sess.run(train_op, feed_dict=pipeline.training_data)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 900, in run
run_metadata_ptr)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1135, in _run
feed_dict_tensor, options, run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1316, in _do_run
run_metadata)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.py", line 1335, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.FailedPreconditionError: Attempting to use uninitialized value metrics/training/accuracy/count
[[Node: metrics/training/accuracy/AssignAdd_1 = AssignAdd[T=DT_FLOAT, use_locking=false, _device="/job:localhost/replica:0/task:0/device:CPU:0"](metrics/training/accuracy/count, metrics/training/accuracy/ToFloat_1)]]

Do you have imagenet pretrained model by your code?
If you have, could you public it?

Thank you very much.

Hello @vonclites,

When I try to run the train_squeezenet.py script, I get following error for slim.deploymet.

Traceback (most recent call last):
File "train_squeezenet.py", line 4, in
from slim.deployment import model_deploy
ImportError: No module named slim.deployment

I am having anaconda environment with python2.7 and Tensorflow 1.6.0 installed using conda environment steps in Tensorflow.

I tried to Google the error but could not found more detail solutions.

Please, let me know about it.

Thanks.

Hello,

For running this network what should be the directory structure for train, val and test folders.
Is it same as what Imagenet has.

Viz

train
|--- folder_1- class#1
|--- folder_2- class#2
test
|--- 15000 images in test folder
val
|--- 50000 images in val folder

Please, let me know.

Thanks.