SSD 受容野について

"""Keras implementation of SSD."""

import keras.backend as K

from keras.layers import Activation

from keras.layers import AtrousConvolution2D

from keras.layers import Convolution2D

from keras.layers import Dense

from keras.layers import Flatten

from keras.layers import GlobalAveragePooling2D

from keras.layers import Input

from keras.layers import MaxPooling2D

from keras.layers import concatenate

from keras.layers import Reshape

from keras.layers import ZeroPadding2D

from keras.models import Model

from ssd_layers import Normalize

from ssd_layers import PriorBox

def SSD300(input_shape, num_classes=21):

"""SSD300 architecture.

# Arguments

input_shape: Shape of the input image,

expected to be either (300, 300, 3) or (3, 300, 300)(not tested).

num_classes: Number of classes including background.

# References

https://arxiv.org/abs/1512.02325

"""

net = {}

# Block 1

input_tensor = input_tensor = Input(shape=input_shape)

img_size = (input_shape[1], input_shape[0])

net['input'] = input_tensor

net['conv1_1'] = Convolution2D(64, 3, 3,

activation='relu',

border_mode='same',

name='conv1_1')(net['input'])

net['conv1_2'] = Convolution2D(64, 3, 3,

activation='relu',

border_mode='same',

name='conv1_2')(net['conv1_1'])

net['pool1'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',

name='pool1')(net['conv1_2'])

# Block 2

net['conv2_1'] = Convolution2D(128, 3, 3,

activation='relu',

border_mode='same',

name='conv2_1')(net['pool1'])

net['conv2_2'] = Convolution2D(128, 3, 3,

activation='relu',

border_mode='same',

name='conv2_2')(net['conv2_1'])

net['pool2'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',

name='pool2')(net['conv2_2'])

# Block 3

net['conv3_1'] = Convolution2D(256, 3, 3,

activation='relu',

border_mode='same',

name='conv3_1')(net['pool2'])

net['conv3_2'] = Convolution2D(256, 3, 3,

activation='relu',

border_mode='same',

name='conv3_2')(net['conv3_1'])

net['conv3_3'] = Convolution2D(256, 3, 3,

activation='relu',

border_mode='same',

name='conv3_3')(net['conv3_2'])

net['pool3'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',

name='pool3')(net['conv3_3'])

# Block 4

net['conv4_1'] = Convolution2D(512, 3, 3,

activation='relu',

border_mode='same',

name='conv4_1')(net['pool3'])

net['conv4_2'] = Convolution2D(512, 3, 3,

activation='relu',

border_mode='same',

name='conv4_2')(net['conv4_1'])

net['conv4_3'] = Convolution2D(512, 3, 3,

activation='relu',

border_mode='same',

name='conv4_3')(net['conv4_2'])

net['pool4'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',

name='pool4')(net['conv4_3'])

# Block 5

net['conv5_1'] = Convolution2D(512, 3, 3,

activation='relu',

border_mode='same',

name='conv5_1')(net['pool4'])

net['conv5_2'] = Convolution2D(512, 3, 3,

activation='relu',

border_mode='same',

name='conv5_2')(net['conv5_1'])

net['conv5_3'] = Convolution2D(512, 3, 3,

activation='relu',

border_mode='same',

name='conv5_3')(net['conv5_2'])

net['pool5'] = MaxPooling2D((3, 3), strides=(1, 1), border_mode='same',

name='pool5')(net['conv5_3'])

# FC6

net['fc6'] = AtrousConvolution2D(1024, 3, 3, atrous_rate=(6, 6),

activation='relu', border_mode='same',

name='fc6')(net['pool5'])

# x = Dropout(0.5, name='drop6')(x)

# FC7

net['fc7'] = Convolution2D(1024, 1, 1, activation='relu',

border_mode='same', name='fc7')(net['fc6'])

# x = Dropout(0.5, name='drop7')(x)

# Block 6

net['conv6_1'] = Convolution2D(256, 1, 1, activation='relu',

border_mode='same',

name='conv6_1')(net['fc7'])

net['conv6_2'] = Convolution2D(512, 3, 3, subsample=(2, 2),

activation='relu', border_mode='same',

name='conv6_2')(net['conv6_1'])

# Block 7

net['conv7_1'] = Convolution2D(128, 1, 1, activation='relu',

border_mode='same',

name='conv7_1')(net['conv6_2'])

net['conv7_2'] = ZeroPadding2D()(net['conv7_1'])

net['conv7_2'] = Convolution2D(256, 3, 3, subsample=(2, 2),

activation='relu', border_mode='valid',

name='conv7_2')(net['conv7_2'])

# Block 8

net['conv8_1'] = Convolution2D(128, 1, 1, activation='relu',

border_mode='same',

name='conv8_1')(net['conv7_2'])

net['conv8_2'] = Convolution2D(256, 3, 3, subsample=(2, 2),

activation='relu', border_mode='same',

name='conv8_2')(net['conv8_1'])

# Last Pool

net['pool6'] = GlobalAveragePooling2D(name='pool6')(net['conv8_2'])

# Prediction from conv4_3

net['conv4_3_norm'] = Normalize(20, name='conv4_3_norm')(net['conv4_3'])

num_priors = 3

x = Convolution2D(num_priors * 4, 3, 3, border_mode='same',

name='conv4_3_norm_mbox_loc')(net['conv4_3_norm'])

net['conv4_3_norm_mbox_loc'] = x

flatten = Flatten(name='conv4_3_norm_mbox_loc_flat')

net['conv4_3_norm_mbox_loc_flat'] = flatten(net['conv4_3_norm_mbox_loc'])

name = 'conv4_3_norm_mbox_conf'

if num_classes != 21:

name += '_{}'.format(num_classes)

x = Convolution2D(num_priors * num_classes, 3, 3, border_mode='same',

name=name)(net['conv4_3_norm'])

net['conv4_3_norm_mbox_conf'] = x

flatten = Flatten(name='conv4_3_norm_mbox_conf_flat')

net['conv4_3_norm_mbox_conf_flat'] = flatten(net['conv4_3_norm_mbox_conf'])

priorbox = PriorBox(img_size, 30.0, aspect_ratios=[2],

variances=[0.1, 0.1, 0.2, 0.2],

name='conv4_3_norm_mbox_priorbox')

net['conv4_3_norm_mbox_priorbox'] = priorbox(net['conv4_3_norm'])

# Prediction from fc7

num_priors = 6

net['fc7_mbox_loc'] = Convolution2D(num_priors * 4, 3, 3,

border_mode='same',

name='fc7_mbox_loc')(net['fc7'])

flatten = Flatten(name='fc7_mbox_loc_flat')

net['fc7_mbox_loc_flat'] = flatten(net['fc7_mbox_loc'])

name = 'fc7_mbox_conf'

if num_classes != 21:

name += '_{}'.format(num_classes)

net['fc7_mbox_conf'] = Convolution2D(num_priors * num_classes, 3, 3,

border_mode='same',

name=name)(net['fc7'])

flatten = Flatten(name='fc7_mbox_conf_flat')

net['fc7_mbox_conf_flat'] = flatten(net['fc7_mbox_conf'])

priorbox = PriorBox(img_size, 60.0, max_size=114.0, aspect_ratios=[2, 3],

variances=[0.1, 0.1, 0.2, 0.2],

name='fc7_mbox_priorbox')

net['fc7_mbox_priorbox'] = priorbox(net['fc7'])

# Prediction from conv6_2

num_priors = 6

x = Convolution2D(num_priors * 4, 3, 3, border_mode='same',

name='conv6_2_mbox_loc')(net['conv6_2'])

net['conv6_2_mbox_loc'] = x

flatten = Flatten(name='conv6_2_mbox_loc_flat')

net['conv6_2_mbox_loc_flat'] = flatten(net['conv6_2_mbox_loc'])

name = 'conv6_2_mbox_conf'

if num_classes != 21:

name += '_{}'.format(num_classes)

x = Convolution2D(num_priors * num_classes, 3, 3, border_mode='same',

name=name)(net['conv6_2'])

net['conv6_2_mbox_conf'] = x

flatten = Flatten(name='conv6_2_mbox_conf_flat')

net['conv6_2_mbox_conf_flat'] = flatten(net['conv6_2_mbox_conf'])

priorbox = PriorBox(img_size, 114.0, max_size=168.0, aspect_ratios=[2, 3],

variances=[0.1, 0.1, 0.2, 0.2],

name='conv6_2_mbox_priorbox')

net['conv6_2_mbox_priorbox'] = priorbox(net['conv6_2'])

# Prediction from conv7_2

num_priors = 6

x = Convolution2D(num_priors * 4, 3, 3, border_mode='same',

name='conv7_2_mbox_loc')(net['conv7_2'])

net['conv7_2_mbox_loc'] = x

flatten = Flatten(name='conv7_2_mbox_loc_flat')

net['conv7_2_mbox_loc_flat'] = flatten(net['conv7_2_mbox_loc'])

name = 'conv7_2_mbox_conf'

if num_classes != 21:

name += '_{}'.format(num_classes)

x = Convolution2D(num_priors * num_classes, 3, 3, border_mode='same',

name=name)(net['conv7_2'])

net['conv7_2_mbox_conf'] = x

flatten = Flatten(name='conv7_2_mbox_conf_flat')

net['conv7_2_mbox_conf_flat'] = flatten(net['conv7_2_mbox_conf'])

priorbox = PriorBox(img_size, 168.0, max_size=222.0, aspect_ratios=[2, 3],

variances=[0.1, 0.1, 0.2, 0.2],

name='conv7_2_mbox_priorbox')

net['conv7_2_mbox_priorbox'] = priorbox(net['conv7_2'])

# Prediction from conv8_2

num_priors = 6

x = Convolution2D(num_priors * 4, 3, 3, border_mode='same',

name='conv8_2_mbox_loc')(net['conv8_2'])

net['conv8_2_mbox_loc'] = x

flatten = Flatten(name='conv8_2_mbox_loc_flat')

net['conv8_2_mbox_loc_flat'] = flatten(net['conv8_2_mbox_loc'])

name = 'conv8_2_mbox_conf'

if num_classes != 21:

name += '_{}'.format(num_classes)

x = Convolution2D(num_priors * num_classes, 3, 3, border_mode='same',

name=name)(net['conv8_2'])

net['conv8_2_mbox_conf'] = x

flatten = Flatten(name='conv8_2_mbox_conf_flat')

net['conv8_2_mbox_conf_flat'] = flatten(net['conv8_2_mbox_conf'])

priorbox = PriorBox(img_size, 222.0, max_size=276.0, aspect_ratios=[2, 3],

variances=[0.1, 0.1, 0.2, 0.2],

name='conv8_2_mbox_priorbox')

net['conv8_2_mbox_priorbox'] = priorbox(net['conv8_2'])

# Prediction from pool6

num_priors = 6

x = Dense(num_priors * 4, name='pool6_mbox_loc_flat')(net['pool6'])

net['pool6_mbox_loc_flat'] = x

name = 'pool6_mbox_conf_flat'

if num_classes != 21:

name += '_{}'.format(num_classes)

x = Dense(num_priors * num_classes, name=name)(net['pool6'])

net['pool6_mbox_conf_flat'] = x

priorbox = PriorBox(img_size, 276.0, max_size=330.0, aspect_ratios=[2, 3],

variances=[0.1, 0.1, 0.2, 0.2],

name='pool6_mbox_priorbox')

if K.image_dim_ordering() == 'tf':

target_shape = (1, 1, 256)

else:

target_shape = (256, 1, 1)

net['pool6_reshaped'] = Reshape(target_shape,

name='pool6_reshaped')(net['pool6'])

net['pool6_mbox_priorbox'] = priorbox(net['pool6_reshaped'])

# Gather all predictions

net['mbox_loc'] = concatenate([net['conv4_3_norm_mbox_loc_flat'],

net['fc7_mbox_loc_flat'],

net['conv6_2_mbox_loc_flat'],

net['conv7_2_mbox_loc_flat'],

net['conv8_2_mbox_loc_flat'],

net['pool6_mbox_loc_flat']],

axis=1, name='mbox_loc')

net['mbox_conf'] = concatenate([net['conv4_3_norm_mbox_conf_flat'],

net['fc7_mbox_conf_flat'],

net['conv6_2_mbox_conf_flat'],

net['conv7_2_mbox_conf_flat'],

net['conv8_2_mbox_conf_flat'],

net['pool6_mbox_conf_flat']],

axis=1, name='mbox_conf')

net['mbox_priorbox'] = concatenate([net['conv4_3_norm_mbox_priorbox'],

net['fc7_mbox_priorbox'],

net['conv6_2_mbox_priorbox'],

net['conv7_2_mbox_priorbox'],

net['conv8_2_mbox_priorbox'],

net['pool6_mbox_priorbox']],

axis=1,

name='mbox_priorbox')

if hasattr(net['mbox_loc'], '_keras_shape'):

num_boxes = net['mbox_loc']._keras_shape[-1] // 4

elif hasattr(net['mbox_loc'], 'int_shape'):

num_boxes = K.int_shape(net['mbox_loc'])[-1] // 4

net['mbox_loc'] = Reshape((num_boxes, 4),

name='mbox_loc_final')(net['mbox_loc'])

net['mbox_conf'] = Reshape((num_boxes, num_classes),

name='mbox_conf_logits')(net['mbox_conf'])

net['mbox_conf'] = Activation('softmax',

name='mbox_conf_final')(net['mbox_conf'])

net['predictions'] = concatenate([net['mbox_loc'],

net['mbox_conf'],

net['mbox_priorbox']],

axis=2,

name='predictions')

model = Model(net['input'], net['predictions'])

return model

 

○目指すこと

・ GTとPBのIoUを増やし、Recall向上

○前提（どんなGTまで対象とするか）

・x方向のみずれたGT

・x,y方向にずれたGT

⇨任意のGTを検出したいため、後者を想定

○方針(どうやってIoUを増やすか)

・PBを大きくする・・・①

・PBをずらして増やす・・・②

・①と②の併用

○トレードオフ

①のメリット

・処理時間が増えにくい（PB数が増えにくい）

①のデメリット

・GT中心が最悪位置のとき、②よりもIoUが小さい

　（PBそのものが大きいため）

※最悪位置：FeatureMap1画素の４隅

②のメリット

・GT中心が最悪位置のとき、①よりもIoUが大きい

②のデメリット

・処理時間が増えやすい（PBが多いため)

①と②共通のリスク

・PBに必要な受容野が広がり、特徴を受容できないリスクがある。

・受容野について

http://joisino.hatenablog.com/entry/2017/07/13/210000