File size: 6,537 Bytes
9a393e2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 |
# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Keras Mask Heads.
Contains Mask prediction head classes for different meta architectures.
All the mask prediction heads have a predict function that receives the
`features` as the first argument and returns `mask_predictions`.
"""
import tensorflow as tf
from object_detection.predictors.heads import head
class ConvolutionalMaskHead(head.KerasHead):
"""Convolutional class prediction head."""
def __init__(self,
is_training,
num_classes,
use_dropout,
dropout_keep_prob,
kernel_size,
num_predictions_per_location,
conv_hyperparams,
freeze_batchnorm,
use_depthwise=False,
mask_height=7,
mask_width=7,
masks_are_class_agnostic=False,
name=None):
"""Constructor.
Args:
is_training: Indicates whether the BoxPredictor is in training mode.
num_classes: Number of classes.
use_dropout: Option to use dropout or not. Note that a single dropout
op is applied here prior to both box and class predictions, which stands
in contrast to the ConvolutionalBoxPredictor below.
dropout_keep_prob: Keep probability for dropout.
This is only used if use_dropout is True.
kernel_size: Size of final convolution kernel. If the
spatial resolution of the feature map is smaller than the kernel size,
then the kernel size is automatically set to be
min(feature_width, feature_height).
num_predictions_per_location: Number of box predictions to be made per
spatial location. Int specifying number of boxes per location.
conv_hyperparams: A `hyperparams_builder.KerasLayerHyperparams` object
containing hyperparameters for convolution ops.
freeze_batchnorm: Bool. Whether to freeze batch norm parameters during
training or not. When training with a small batch size (e.g. 1), it is
desirable to freeze batch norm update and use pretrained batch norm
params.
use_depthwise: Whether to use depthwise convolutions for prediction
steps. Default is False.
mask_height: Desired output mask height. The default value is 7.
mask_width: Desired output mask width. The default value is 7.
masks_are_class_agnostic: Boolean determining if the mask-head is
class-agnostic or not.
name: A string name scope to assign to the model. If `None`, Keras
will auto-generate one from the class name.
Raises:
ValueError: if min_depth > max_depth.
"""
super(ConvolutionalMaskHead, self).__init__(name=name)
self._is_training = is_training
self._num_classes = num_classes
self._use_dropout = use_dropout
self._dropout_keep_prob = dropout_keep_prob
self._kernel_size = kernel_size
self._num_predictions_per_location = num_predictions_per_location
self._use_depthwise = use_depthwise
self._mask_height = mask_height
self._mask_width = mask_width
self._masks_are_class_agnostic = masks_are_class_agnostic
self._mask_predictor_layers = []
# Add a slot for the background class.
if self._masks_are_class_agnostic:
self._num_masks = 1
else:
self._num_masks = self._num_classes
num_mask_channels = self._num_masks * self._mask_height * self._mask_width
if self._use_dropout:
self._mask_predictor_layers.append(
# The Dropout layer's `training` parameter for the call method must
# be set implicitly by the Keras set_learning_phase. The object
# detection training code takes care of this.
tf.keras.layers.Dropout(rate=1.0 - self._dropout_keep_prob))
if self._use_depthwise:
self._mask_predictor_layers.append(
tf.keras.layers.DepthwiseConv2D(
[self._kernel_size, self._kernel_size],
padding='SAME',
depth_multiplier=1,
strides=1,
dilation_rate=1,
name='MaskPredictor_depthwise',
**conv_hyperparams.params()))
self._mask_predictor_layers.append(
conv_hyperparams.build_batch_norm(
training=(is_training and not freeze_batchnorm),
name='MaskPredictor_depthwise_batchnorm'))
self._mask_predictor_layers.append(
conv_hyperparams.build_activation_layer(
name='MaskPredictor_depthwise_activation'))
self._mask_predictor_layers.append(
tf.keras.layers.Conv2D(
num_predictions_per_location * num_mask_channels, [1, 1],
name='MaskPredictor',
**conv_hyperparams.params(use_bias=True)))
else:
self._mask_predictor_layers.append(
tf.keras.layers.Conv2D(
num_predictions_per_location * num_mask_channels,
[self._kernel_size, self._kernel_size],
padding='SAME',
name='MaskPredictor',
**conv_hyperparams.params(use_bias=True)))
def _predict(self, features):
"""Predicts boxes.
Args:
features: A float tensor of shape [batch_size, height, width, channels]
containing image features.
Returns:
mask_predictions: A float tensors of shape
[batch_size, num_anchors, num_masks, mask_height, mask_width]
representing the mask predictions for the proposals.
"""
mask_predictions = features
for layer in self._mask_predictor_layers:
mask_predictions = layer(mask_predictions)
batch_size = features.get_shape().as_list()[0]
if batch_size is None:
batch_size = tf.shape(features)[0]
mask_predictions = tf.reshape(
mask_predictions,
[batch_size, -1, self._num_masks, self._mask_height, self._mask_width])
return mask_predictions
|