diff --git a/keras_contrib/applications/nasnet.py b/keras_contrib/applications/nasnet.py index 12895a6..ccb494a 100644 --- a/keras_contrib/applications/nasnet.py +++ b/keras_contrib/applications/nasnet.py @@ -44,13 +44,22 @@ from keras import backend as K _BN_DECAY = 0.9997 _BN_EPSILON = 1e-3 +NASNET_MOBILE_WEIGHT_PATH = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.0/NASNet-mobile.h5" +NASNET_MOBILE_WEIGHT_PATH_NO_TOP = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.0/NASNet-mobile-no-top.h5" +NASNET_MOBILE_WEIGHT_PATH_WITH_AUXULARY = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.0/NASNet-auxiliary-mobile.h5" +NASNET_MOBILE_WEIGHT_PATH_WITH_AUXULARY_NO_TOP = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.0/NASNet-auxiliary-mobile-no-top.h5" +NASNET_LARGE_WEIGHT_PATH = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.1/NASNet-large.h5" +NASNET_LARGE_WEIGHT_PATH_NO_TOP = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.1/NASNet-large-no-top.h5" +NASNET_LARGE_WEIGHT_PATH_WITH_auxiliary = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.1/NASNet-auxiliary-large.h5" +NASNET_LARGE_WEIGHT_PATH_WITH_auxiliary_NO_TOP = "https://github.com/titu1994/Keras-NASNet/releases/download/v1.1/NASNet-auxiliary-large-no-top.h5" + def NASNet(input_shape=None, penultimate_filters=4032, nb_blocks=6, stem_filters=96, skip_reduction=True, - use_auxilary_branch=False, + use_auxiliary_branch=False, filters_multiplier=2, dropout=0.5, weight_decay=5e-5, @@ -85,7 +94,7 @@ def NASNet(input_shape=None, stem_filters: number of filters in the initial stem block skip_reduction: Whether to skip the reduction step at the tail end of the network. Set to `False` for CIFAR models. - use_auxilary_branch: Whether to use the auxilary branch during + use_auxiliary_branch: Whether to use the auxiliary branch during training or evaluation. filters_multiplier: controls the width of the network. - If `filters_multiplier` < 1.0, proportionally decreases the number @@ -205,16 +214,16 @@ def NASNet(input_shape=None, for i in range(nb_blocks): x, p = _normal_A(x, p, filters * filters_multiplier, weight_decay, id='%d' % (nb_blocks + i + 1)) - auxilary_x = None + auxiliary_x = None if not skip_reduction: # imagenet / mobile mode - if use_auxilary_branch: - auxilary_x = _add_auxilary_head(x, classes, weight_decay) + if use_auxiliary_branch: + auxiliary_x = _add_auxiliary_head(x, classes, weight_decay) x, p0 = _reduction_A(x, p, filters * filters_multiplier ** 2, weight_decay, id='reduce_%d' % (2 * nb_blocks)) if skip_reduction: # CIFAR mode - if use_auxilary_branch: - auxilary_x = _add_auxilary_head(x, classes, weight_decay) + if use_auxiliary_branch: + auxiliary_x = _add_auxiliary_head(x, classes, weight_decay) p = p0 if not skip_reduction else p @@ -241,14 +250,53 @@ def NASNet(input_shape=None, inputs = img_input # Create model. - if use_auxilary_branch: - model = Model(inputs, [x, auxilary_x], name='NASNet_with_auxilary') + if use_auxiliary_branch: + model = Model(inputs, [x, auxiliary_x], name='NASNet_with_auxiliary') else: model = Model(inputs, x, name='NASNet') - # load weights (when available) - if weights is not None: - warnings.warn('Weights of NASNet models have not yet been ported to Keras') + # load weights + if weights == 'imagenet': + if default_size == 224: # mobile version + if include_top: + if use_auxiliary_branch: + weight_path = NASNET_MOBILE_WEIGHT_PATH_WITH_AUXULARY + model_name = 'nasnet_mobile_with_aux.h5' + else: + weight_path = NASNET_MOBILE_WEIGHT_PATH + model_name = 'nasnet_mobile.h5' + else: + if use_auxiliary_branch: + weight_path = NASNET_MOBILE_WEIGHT_PATH_WITH_AUXULARY_NO_TOP + model_name = 'nasnet_mobile_with_aux_no_top.h5' + else: + weight_path = NASNET_MOBILE_WEIGHT_PATH_NO_TOP + model_name = 'nasnet_mobile_no_top.h5' + + weights_file = get_file(model_name, weight_path, cache_subdir='models') + model.load_weights(weights_file, by_name=True) + + elif default_size == 331: # large version + if include_top: + if use_auxiliary_branch: + weight_path = NASNET_LARGE_WEIGHT_PATH_WITH_auxiliary + model_name = 'nasnet_large_with_aux.h5' + else: + weight_path = NASNET_LARGE_WEIGHT_PATH + model_name = 'nasnet_large.h5' + else: + if use_auxiliary_branch: + weight_path = NASNET_LARGE_WEIGHT_PATH_WITH_auxiliary_NO_TOP + model_name = 'nasnet_large_with_aux_no_top.h5' + else: + weight_path = NASNET_LARGE_WEIGHT_PATH_NO_TOP + model_name = 'nasnet_large_no_top.h5' + + weights_file = get_file(model_name, weight_path, cache_subdir='models') + model.load_weights(weights_file, by_name=True) + + else: + raise ValueError('ImageNet weights can only be loaded on NASNetLarge or NASNetMobile') if old_data_format: K.set_image_data_format(old_data_format) @@ -259,9 +307,9 @@ def NASNet(input_shape=None, def NASNetLarge(input_shape=(331, 331, 3), dropout=0.5, weight_decay=5e-5, - use_auxilary_branch=False, + use_auxiliary_branch=False, include_top=True, - weights=None, + weights='imagenet', input_tensor=None, pooling=None, classes=1000): @@ -278,7 +326,7 @@ def NASNetLarge(input_shape=(331, 331, 3), It should have exactly 3 inputs channels, and width and height should be no smaller than 32. E.g. `(224, 224, 3)` would be one valid value. - use_auxilary_branch: Whether to use the auxilary branch during + use_auxiliary_branch: Whether to use the auxiliary branch during training or evaluation. dropout: dropout rate weight_decay: l2 regularization weight @@ -321,8 +369,8 @@ def NASNetLarge(input_shape=(331, 331, 3), penultimate_filters=4032, nb_blocks=6, stem_filters=96, - skip_reduction=True, - use_auxilary_branch=use_auxilary_branch, + skip_reduction=False, + use_auxiliary_branch=use_auxiliary_branch, filters_multiplier=2, dropout=dropout, weight_decay=weight_decay, @@ -337,9 +385,9 @@ def NASNetLarge(input_shape=(331, 331, 3), def NASNetMobile(input_shape=(224, 224, 3), dropout=0.5, weight_decay=4e-5, - use_auxilary_branch=False, + use_auxiliary_branch=False, include_top=True, - weights=None, + weights='imagenet', input_tensor=None, pooling=None, classes=1000): @@ -356,7 +404,7 @@ def NASNetMobile(input_shape=(224, 224, 3), It should have exactly 3 inputs channels, and width and height should be no smaller than 32. E.g. `(224, 224, 3)` would be one valid value. - use_auxilary_branch: Whether to use the auxilary branch during + use_auxiliary_branch: Whether to use the auxiliary branch during training or evaluation. dropout: dropout rate weight_decay: l2 regularization weight @@ -400,7 +448,7 @@ def NASNetMobile(input_shape=(224, 224, 3), nb_blocks=4, stem_filters=32, skip_reduction=False, - use_auxilary_branch=use_auxilary_branch, + use_auxiliary_branch=use_auxiliary_branch, filters_multiplier=2, dropout=dropout, weight_decay=weight_decay, @@ -415,7 +463,7 @@ def NASNetMobile(input_shape=(224, 224, 3), def NASNetCIFAR(input_shape=(32, 32, 3), dropout=0.0, weight_decay=5e-4, - use_auxilary_branch=False, + use_auxiliary_branch=False, include_top=True, weights=None, input_tensor=None, @@ -434,7 +482,7 @@ def NASNetCIFAR(input_shape=(32, 32, 3), It should have exactly 3 inputs channels, and width and height should be no smaller than 32. E.g. `(32, 32, 3)` would be one valid value. - use_auxilary_branch: Whether to use the auxilary branch during + use_auxiliary_branch: Whether to use the auxiliary branch during training or evaluation. dropout: dropout rate weight_decay: l2 regularization weight @@ -476,9 +524,9 @@ def NASNetCIFAR(input_shape=(32, 32, 3), return NASNet(input_shape, penultimate_filters=768, nb_blocks=6, - stem_filters=96, + stem_filters=32, skip_reduction=True, - use_auxilary_branch=use_auxilary_branch, + use_auxiliary_branch=use_auxiliary_branch, filters_multiplier=2, dropout=dropout, weight_decay=weight_decay, @@ -545,7 +593,7 @@ def _adjust_block(p, ip, filters, weight_decay=5e-5, id=None): if p is None: p = ip - if p._keras_shape[img_dim] != ip._keras_shape[img_dim]: + elif p._keras_shape[img_dim] != ip._keras_shape[img_dim]: with K.name_scope('adjust_reduction_block_%s' % id): p = Activation('relu', name='adjust_relu_1_%s' % id)(p) @@ -598,13 +646,14 @@ def _normal_A(ip, p, filters, weight_decay=5e-5, id=None): name='normal_bn_1_%s' % id)(h) with K.name_scope('block_1'): - x1 = _separable_conv_block(h, filters, weight_decay=weight_decay, id='normal_left1_%s' % id) - x1 = add([x1, h], name='normal_add_1_%s' % id) + x1_1 = _separable_conv_block(h, filters, kernel_size=(5, 5), weight_decay=weight_decay, + id='normal_left1_%s' % id) + x1_2 = _separable_conv_block(p, filters, weight_decay=weight_decay, id='normal_right1_%s' % id) + x1 = add([x1_1, x1_2], name='normal_add_1_%s' % id) with K.name_scope('block_2'): - x2_1 = _separable_conv_block(p, filters, weight_decay=weight_decay, id='normal_left2_%s' % id) - x2_2 = _separable_conv_block(h, filters, kernel_size=(5, 5), weight_decay=weight_decay, - id='normal_right2_%s' % id) + x2_1 = _separable_conv_block(p, filters, (5, 5), weight_decay=weight_decay, id='normal_left2_%s' % id) + x2_2 = _separable_conv_block(p, filters, (3, 3), weight_decay=weight_decay, id='normal_right2_%s' % id) x2 = add([x2_1, x2_2], name='normal_add_2_%s' % id) with K.name_scope('block_3'): @@ -617,11 +666,10 @@ def _normal_A(ip, p, filters, weight_decay=5e-5, id=None): x4 = add([x4_1, x4_2], name='normal_add_4_%s' % id) with K.name_scope('block_5'): - x5_1 = _separable_conv_block(p, filters, (5, 5), weight_decay=weight_decay, id='normal_left5_%s' % id) - x5_2 = _separable_conv_block(p, filters, (3, 3), weight_decay=weight_decay, id='normal_right5_%s' % id) - x5 = add([x5_1, x5_2], name='normal_add_5_%s' % id) + x5 = _separable_conv_block(h, filters, weight_decay=weight_decay, id='normal_left5_%s' % id) + x5 = add([x5, h], name='normal_add_5_%s' % id) - x = concatenate([p, x2, x5, x3, x4, x1], axis=channel_dim, name='normal_concat_%s' % id) + x = concatenate([p, x1, x2, x3, x4, x5], axis=channel_dim, name='normal_concat_%s' % id) return x, ip @@ -651,10 +699,10 @@ def _reduction_A(ip, p, filters, weight_decay=5e-5, id=None): name='reduction_bn_1_%s' % id)(h) with K.name_scope('block_1'): - x1_1 = _separable_conv_block(p, filters, (7, 7), strides=(2, 2), weight_decay=weight_decay, + x1_1 = _separable_conv_block(h, filters, (5, 5), strides=(2, 2), weight_decay=weight_decay, id='reduction_left1_%s' % id) - x1_2 = _separable_conv_block(h, filters, (5, 5), strides=(2, 2), weight_decay=weight_decay, - id='reduction_right1_%s' % id) + x1_2 = _separable_conv_block(p, filters, (7, 7), strides=(2, 2), weight_decay=weight_decay, + id='reduction_1_%s' % id) x1 = add([x1_1, x1_2], name='reduction_add_1_%s' % id) with K.name_scope('block_2'): @@ -670,22 +718,23 @@ def _reduction_A(ip, p, filters, weight_decay=5e-5, id=None): x3 = add([x3_1, x3_2], name='reduction_add3_%s' % id) with K.name_scope('block_4'): - x4_1 = MaxPooling2D((3, 3), strides=(2, 2), padding='same', name='reduction_left4_%s' % id)(h) - x4_2 = _separable_conv_block(x1, filters, (3, 3), weight_decay=weight_decay, id='reduction_right4_%s' % id) - x4 = add([x4_1, x4_2], name='reduction_add4_%s' % id) + x4 = AveragePooling2D((3, 3), strides=(1, 1), padding='same', name='reduction_left4_%s' % id)(x1) + x4 = add([x2, x4]) with K.name_scope('block_5'): - x5 = AveragePooling2D((3, 3), strides=(1, 1), padding='same', name='reduction_left5_%s' % id)(x1) + x5_1 = _separable_conv_block(x1, filters, (3, 3), weight_decay=weight_decay, id='reduction_left4_%s' % id) + x5_2 = MaxPooling2D((3, 3), strides=(2, 2), padding='same', name='reduction_right5_%s' % id)(h) + x5 = add([x5_1, x5_2], name='reduction_add4_%s' % id) - x = concatenate([x2, x3, x5, x4], axis=channel_dim, name='reduction_concat_%s' % id) + x = concatenate([x2, x3, x4, x5], axis=channel_dim, name='reduction_concat_%s' % id) return x, ip -def _add_auxilary_head(x, classes, weight_decay): - '''Adds an auxilary head for training the model +def _add_auxiliary_head(x, classes, weight_decay): + '''Adds an auxiliary head for training the model From section A.7 "Training of ImageNet models" of the paper, all NASNet models are - trained using an auxilary classifier around 2/3 of the depth of the network, with + trained using an auxiliary classifier around 2/3 of the depth of the network, with a loss weight of 0.4 # Arguments @@ -700,23 +749,23 @@ def _add_auxilary_head(x, classes, weight_decay): img_width = 2 if K.image_data_format() == 'channels_last' else 3 channel_axis = 1 if K.image_data_format() == 'channels_first' else -1 - with K.name_scope('auxilary_branch'): - auxilary_x = Activation('relu')(x) - auxilary_x = AveragePooling2D((5, 5), strides=(3, 3), padding='valid', name='aux_pool')(auxilary_x) - auxilary_x = Conv2D(128, (1, 1), padding='same', use_bias=False, name='aux_conv_projection', - kernel_initializer='he_normal', kernel_regularizer=l2(weight_decay))(auxilary_x) - auxilary_x = BatchNormalization(axis=channel_axis, momentum=_BN_DECAY, epsilon=_BN_EPSILON, - name='aux_bn_projection')(auxilary_x) - auxilary_x = Activation('relu')(auxilary_x) + with K.name_scope('auxiliary_branch'): + auxiliary_x = Activation('relu')(x) + auxiliary_x = AveragePooling2D((5, 5), strides=(3, 3), padding='valid', name='aux_pool')(auxiliary_x) + auxiliary_x = Conv2D(128, (1, 1), padding='same', use_bias=False, name='aux_conv_projection', + kernel_initializer='he_normal', kernel_regularizer=l2(weight_decay))(auxiliary_x) + auxiliary_x = BatchNormalization(axis=channel_axis, momentum=_BN_DECAY, epsilon=_BN_EPSILON, + name='aux_bn_projection')(auxiliary_x) + auxiliary_x = Activation('relu')(auxiliary_x) - auxilary_x = Conv2D(768, (auxilary_x._keras_shape[img_height], auxilary_x._keras_shape[img_width]), - padding='valid', use_bias=False, kernel_initializer='he_normal', - kernel_regularizer=l2(weight_decay), name='aux_conv_reduction')(auxilary_x) - auxilary_x = BatchNormalization(axis=channel_axis, momentum=_BN_DECAY, epsilon=_BN_EPSILON, - name='aux_bn_reduction')(auxilary_x) - auxilary_x = Activation('relu')(auxilary_x) + auxiliary_x = Conv2D(768, (auxiliary_x._keras_shape[img_height], auxiliary_x._keras_shape[img_width]), + padding='valid', use_bias=False, kernel_initializer='he_normal', + kernel_regularizer=l2(weight_decay), name='aux_conv_reduction')(auxiliary_x) + auxiliary_x = BatchNormalization(axis=channel_axis, momentum=_BN_DECAY, epsilon=_BN_EPSILON, + name='aux_bn_reduction')(auxiliary_x) + auxiliary_x = Activation('relu')(auxiliary_x) - auxilary_x = GlobalAveragePooling2D()(auxilary_x) - auxilary_x = Dense(classes, activation='softmax', kernel_regularizer=l2(weight_decay), - name='aux_predictions')(auxilary_x) - return auxilary_x + auxiliary_x = GlobalAveragePooling2D()(auxiliary_x) + auxiliary_x = Dense(classes, activation='softmax', kernel_regularizer=l2(weight_decay), + name='aux_predictions')(auxiliary_x) + return auxiliary_x