guillaumemougeot/dogfacenet

FaceNet implementation for dog identification

dogfacenet's Issues

run error

I ran command python ./dogfacenet/dogfacenet.py with environment ubuntu 18.04.2 LTS, python 3.6.8, and tensorflow 1.13.1.

And I got this error. I don't have any idea to fix this error and why this error show up.

Traceback (most recent call last):
File "./dogfacenet/dogfacenet.py", line 219, in
validation_steps=VALIDATION_STEPS)]
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 1426, in fit_generator
initial_epoch=initial_epoch)
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/engine/training_generator.py", line 177, in model_iteration
batch_data = _get_next_batch(output_generator, mode)
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/engine/training_generator.py", line 258, in _get_next_batch
generator_output = next(output_generator)
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/utils/data_utils.py", line 767, in get
six.reraise(*sys.exc_info())
File "/home/aji3rang/.local/lib/python3.6/site-packages/six.py", line 696, in reraise
raise value
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/utils/data_utils.py", line 743, in get
inputs = self.queue.get(block=True).get()
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/multiprocessing/pool.py", line 644, in get
raise self._value
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/multiprocessing/pool.py", line 119, in worker
result = (True, func(*args, **kwds))
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/utils/data_utils.py", line 680, in next_sample
return six.next(_SHARED_SEQUENCES[uid])
File "/home/aji3rang/dognet/DogFaceNet/dogfacenet/online_training.py", line 329, in online_adaptive_hard_image_generator
steps=int(np.ceil(len(subfilenames)/32)))
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 1536, in predict_generator
verbose=verbose)
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/engine/training_generator.py", line 171, in model_iteration
model.reset_metrics()
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 1119, in reset_metrics
m.reset_states()
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/metrics.py", line 460, in reset_states
K.set_value(v, 0)
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/keras/backend.py", line 2847, in set_value
get_session().run(assign_op, feed_dict={assign_placeholder: value})
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 929, in run
run_metadata_ptr)
File "/home/aji3rang/anaconda3/envs/tf3/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1077, in _run
raise RuntimeError('The Session graph is empty. Add operations to the '
RuntimeError: The Session graph is empty. Add operations to the graph before calling run().

can you share the dataset after alignment

I read your paper and feel really excited because it is near to the problem I am facing with.

Would you please to share the dataset after preprocessing (the dataset on the right of the image)

Where is landmark data?

I like your awesome project.

I download the dataset below page.
https://github.com/GuillaumeMougeot/DogFaceNet/releases/

but I can't find landmark data or a pre-train model.
It is only 244 * 244 images.
Can you share it?

Could you share high resolution images?

Can you share the pre-train model?

I can't find landmark data or a pre-train model.
Can you share it?

Bad results with eval.py

During training a get val_triplet_acc 85-90 % and when i use the eval.py the prediction are random ( 48-52 % accuracy). i don't understand how this is possible?. I am using the lfw dataset