diff --git a/bob/learn/tensorflow/test/test_onegraph_model_fn.py b/bob/learn/tensorflow/test/test_onegraph_model_fn.py
index c6f4d54f81a03ea6e8ad219a197a34f1e9381ed2..fadf49dc9725ce24956f8e4d31c2577922b3ec2e 100755
--- a/bob/learn/tensorflow/test/test_onegraph_model_fn.py
+++ b/bob/learn/tensorflow/test/test_onegraph_model_fn.py
@@ -63,28 +63,28 @@ def run_cnn(embedding_validation):
optimizer=tf.train.GradientDescentOptimizer(learning_rate),
n_classes=10,
loss_op=mean_cross_entropy_loss,
- embedding_validation=embedding_validation)
+ embedding_validation=embedding_validation,
+ validation_batch_size=validation_batch_size
+ )
- data, labels = shuffle_data_and_labels([tfrecord_train], data_shape, data_type, batch_size, epochs=epochs)
def input_fn():
return shuffle_data_and_labels(tfrecord_train, data_shape, data_type,
- batch_size, epochs=epochs)
+ batch_size, epochs=epochs)
def input_fn_validation():
return batch_data_and_labels(tfrecord_validation, data_shape, data_type,
validation_batch_size, epochs=epochs)
hooks = [LoggerHookEstimator(trainer, 16, 100)]
- trainer.train(input_fn, steps=steps, hooks = hooks)
+ trainer.train(input_fn, steps=steps, hooks=hooks)
- # TODO: REMOVE THIS HACK
if not embedding_validation:
acc = trainer.evaluate(input_fn_validation)
- assert acc > 0.80
+ assert acc['accuracy'] > 0.80
else:
- assert True
-
-
+ acc = trainer.evaluate(input_fn_validation)
+ assert acc['accuracy'] > 0.80
+
finally:
try:
os.unlink(tfrecord_train)
diff --git a/bob/learn/tensorflow/test/test_utils.py b/bob/learn/tensorflow/test/test_utils.py
index 966fd66c0036ce00feba0b9c82d411b8615b0f26..23f3ded505b260b9cf8f7cb3d3e224fb8853bb17 100755
--- a/bob/learn/tensorflow/test/test_utils.py
+++ b/bob/learn/tensorflow/test/test_utils.py
@@ -3,7 +3,8 @@
# @author: Tiago de Freitas Pereira <tiago.pereira@idiap.ch>
import numpy
-from bob.learn.tensorflow.utils import compute_embedding_accuracy, cdist, compute_embedding_accuracy_tensors
+from bob.learn.tensorflow.utils import compute_embedding_accuracy, cdist,\
+ compute_embedding_accuracy_tensors, predict_using_tensors
import tensorflow as tf
@@ -11,6 +12,7 @@ import tensorflow as tf
Some unit tests for the datashuffler
"""
+
def test_embedding_accuracy():
numpy.random.seed(10)
@@ -53,8 +55,7 @@ def test_embedding_accuracy_tensors():
data = tf.convert_to_tensor(data.astype("float32"))
labels = tf.convert_to_tensor(labels.astype("int64"))
-
+
sess = tf.Session()
accuracy = sess.run(compute_embedding_accuracy_tensors(data, labels))
assert accuracy == 1.
-
diff --git a/bob/learn/tensorflow/trainers/LogitsTrainer.py b/bob/learn/tensorflow/trainers/LogitsTrainer.py
index c4b63d24378e48c51420c8c36c534d32b354c226..7204396701590e98779e61d0f88b217641c8cf62 100755
--- a/bob/learn/tensorflow/trainers/LogitsTrainer.py
+++ b/bob/learn/tensorflow/trainers/LogitsTrainer.py
@@ -15,7 +15,7 @@ import time
from bob.learn.tensorflow.network.utils import append_logits
from tensorflow.python.estimator import estimator
from bob.learn.tensorflow.utils import reproducible
-from bob.learn.tensorflow.utils import compute_embedding_accuracy_tensors
+from bob.learn.tensorflow.utils import predict_using_tensors
import logging
@@ -48,6 +48,10 @@ class LogitsTrainer(estimator.Estimator):
model_dir:
Model path
+
+ validation_batch_size:
+ Size of the batch for validation. This value is used when the
+ validation with embeddings is used. This is a hack.
"""
def __init__(self,
@@ -58,6 +62,7 @@ class LogitsTrainer(estimator.Estimator):
loss_op=None,
embedding_validation=False,
model_dir="",
+ validation_batch_size=None,
):
self.architecture = architecture
@@ -79,7 +84,6 @@ class LogitsTrainer(estimator.Estimator):
if self.n_classes <=0:
raise ValueError("Number of classes must be greated than 0")
-
def _model_fn(features, labels, mode, params, config):
# Building one graph
@@ -90,7 +94,7 @@ class LogitsTrainer(estimator.Estimator):
# Compute the embeddings
embeddings = tf.nn.l2_normalize(prelogits, 1)
predictions = {
- "embeddings":embeddings
+ "embeddings": embeddings
}
else:
@@ -107,7 +111,6 @@ class LogitsTrainer(estimator.Estimator):
# Calculate Loss (for both TRAIN and EVAL modes)
self.loss = self.loss_op(logits, labels)
-
# Configure the Training Op (for TRAIN mode)
if mode == tf.estimator.ModeKeys.TRAIN:
@@ -117,8 +120,8 @@ class LogitsTrainer(estimator.Estimator):
train_op=train_op)
if self.embedding_validation:
- #eval_metric_ops = {"accuracy": compute_embedding_accuracy_tensors(predictions["embeddings"], labels)}
- eval_metric_ops = {} # TODO: I still don't know how to compute this with an unknown size
+ predictions_op = predict_using_tensors(predictions["embeddings"], labels, num=validation_batch_size)
+ eval_metric_ops = {"accuracy": tf.metrics.accuracy(labels=labels, predictions=predictions_op)}
return tf.estimator.EstimatorSpec(mode=mode, loss=self.loss, eval_metric_ops=eval_metric_ops)
else:
diff --git a/bob/learn/tensorflow/utils/util.py b/bob/learn/tensorflow/utils/util.py
index 7aae42fa20e19b5aebad1748d809350ec5156e4c..d3a9b9943e82907629491205f0407cd149e43437 100755
--- a/bob/learn/tensorflow/utils/util.py
+++ b/bob/learn/tensorflow/utils/util.py
@@ -46,7 +46,6 @@ def load_mnist(perc_train=0.9):
return train_data, train_labels, validation_data, validation_labels
-
def create_mnist_tfrecord(tfrecords_filename, data, labels, n_samples=6000):
def _bytes_feature(value):
@@ -150,43 +149,59 @@ def debug_embbeding(image, architecture, embbeding_dim=2, feature_layer="fc3"):
embeddings[i] = embedding
return embeddings
-
def cdist(A):
+ """
+ Compute a pairwise euclidean distance in the same fashion
+ as in scipy.spation.distance.cdist
+ """
with tf.variable_scope('Pairwisedistance'):
+ #ones_1 = tf.ones(shape=(1, A.shape.as_list()[0]))
+ ones_1 = tf.reshape(tf.cast(tf.ones_like(A), tf.float32)[:, 0], [1, -1])
p1 = tf.matmul(
tf.expand_dims(tf.reduce_sum(tf.square(A), 1), 1),
- tf.ones(shape=(1, A.shape.as_list()[0]))
+ ones_1
)
+
+ #ones_2 = tf.ones(shape=(A.shape.as_list()[0], 1))
+ ones_2 = tf.reshape(tf.cast(tf.ones_like(A), tf.float32)[:, 0], [-1, 1])
p2 = tf.transpose(tf.matmul(
tf.reshape(tf.reduce_sum(tf.square(A), 1), shape=[-1, 1]),
- tf.ones(shape=(A.shape.as_list()[0], 1)),
+ ones_2,
transpose_b=True
))
return tf.sqrt(tf.add(p1, p2) - 2 * tf.matmul(A, A, transpose_b=True))
-def compute_embedding_accuracy_tensors(embedding, labels):
+def predict_using_tensors(embedding, labels, num=None):
"""
- Compute the accuracy through exhaustive comparisons between the embeddings using tensors
+ Compute the predictions through exhaustive comparisons between
+ embeddings using tensors
"""
-
- distances = cdist(embedding)
# Fitting the main diagonal with infs (removing comparisons with the same sample)
- inf = numpy.ones(10)*numpy.inf
- inf = inf.astype("float32")
+ inf = tf.cast(tf.ones_like(labels), tf.float32) * numpy.inf
distances = cdist(embedding)
distances = tf.matrix_set_diag(distances, inf)
indexes = tf.argmin(distances, axis=1)
+ return [labels[i] for i in tf.unstack(indexes, num=num)]
+
+
+def compute_embedding_accuracy_tensors(embedding, labels, num=None):
+ """
+ Compute the accuracy through exhaustive comparisons between the embeddings using tensors
+ """
+
+ # Fitting the main diagonal with infs (removing comparisons with the same sample)
+ predictions = predict_using_tensors(embedding, labels, num=num)
+ matching = [tf.equal(p, l) for p, l in zip(tf.unstack(predictions, num=num), tf.unstack(labels, num=num))]
+
+ return tf.reduce_sum(tf.cast(matching, tf.uint8))/len(predictions)
- matching = [ tf.equal(labels[i],labels[j]) for i,j in zip(range(indexes.get_shape().as_list()[0]), tf.unstack(indexes))]
- return tf.reduce_sum(tf.cast(matching, tf.uint8))/indexes.get_shape().as_list()[0]
-
def compute_embedding_accuracy(embedding, labels):
"""
Compute the accuracy through exhaustive comparisons between the embeddings