Tensorflow2上のKerasでSSDを実行を試したときのメモ
※ 2021.05.21 kerasをtensorflow.kerasに変更
Kerasを使ってSSDを実行してみようと思ったら、Tensorflow1.x上の情報ばかり出てきて、Tensorflow2で実行するのに手間取ったのでメモ。
(最終的にopenVINOに持っていきたいなぁ、と思って試し始めたんだけど、openVINOにサクッと変換できるのはcaffeだった😅)
ということで、やったけど 続きはないかも。実行遅いし…😩💨
【実験】学習済のSSDを使ってPaPeRo i に複数種類の物体を数えさせる を参考に、KerasをTensorflow2で動かしてみる。
Pythonは3.8を使用した(念のため明記しておく)。
Anaconda使ってないし、Paperoじゃないので、そのまんまじゃないけど、
とりあえずSSDを動かしてみよう。
まずは、python仮想環境の準備をする。
pyenv環境前提。
# ベースディレクトリとpython仮想環境の準備
mkdir -p /work/keras2/
cd /work/keras2/
pyenv virtualenv 3.8.8 keras2
pyenv local keras2
pip install --upgrade pip setuptools
# pythonモジュールのインストール
pip install opencv-python
pip install tensorflow
# kerasはtensorflow内を直接参照にしたので不要
# pip install keras
pip install matplotlib
pip install imageio
インストールされているpythonモジュールは以下の通り
absl-py==0.12.0
astunparse==1.6.3
cachetools==4.2.2
certifi==2020.12.5
chardet==4.0.0
cycler==0.10.0
flatbuffers==1.12
gast==0.4.0
google-auth==1.30.0
google-auth-oauthlib==0.4.4
google-pasta==0.2.0
grpcio==1.34.1
h5py==3.1.0
idna==2.10
imageio==2.9.0
keras-nightly==2.5.0.dev2021032900
Keras-Preprocessing==1.1.2
kiwisolver==1.3.1
Markdown==3.3.4
matplotlib==3.4.2
numpy==1.19.5
oauthlib==3.1.0
opencv-python==4.5.2.52
opt-einsum==3.3.0
Pillow==8.2.0
protobuf==3.17.0
pyasn1==0.4.8
pyasn1-modules==0.2.8
pyparsing==2.4.7
python-dateutil==2.8.1
requests==2.25.1
requests-oauthlib==1.3.0
rsa==4.7.2
six==1.15.0
tensorboard==2.5.0
tensorboard-data-server==0.6.1
tensorboard-plugin-wit==1.8.0
tensorflow==2.5.0
tensorflow-estimator==2.5.0
termcolor==1.1.0
typing-extensions==3.7.4.3
urllib3==1.26.4
Werkzeug==2.0.1
wrapt==1.12.1
[!NOTE] Keras修正以前の状態は以下
absl-py==0.12.0 astunparse==1.6.3 cachetools==4.2.1 certifi==2020.12.5 chardet==4.0.0 cycler==0.10.0 flatbuffers==1.12 gast==0.3.3 google-auth==1.28.0 google-auth-oauthlib==0.4.3 google-pasta==0.2.0 grpcio==1.32.0 h5py==2.10.0 idna==2.10 imageio==2.9.0 Keras==2.4.3 Keras-Preprocessing==1.1.2 kiwisolver==1.3.1 Markdown==3.3.4 matplotlib==3.3.4 numpy==1.19.5 oauthlib==3.1.0 opencv-python==4.5.1.48 opt-einsum==3.3.0 Pillow==8.1.2 protobuf==3.15.6 pyasn1==0.4.8 pyasn1-modules==0.2.8 pyparsing==2.4.7 python-dateutil==2.8.1 PyYAML==5.4.1 requests==2.25.1 requests-oauthlib==1.3.0 rsa==4.7.2 scipy==1.6.1 six==1.15.0 tensorboard==2.4.1 tensorboard-plugin-wit==1.8.0 tensorflow==2.4.1 tensorflow-estimator==2.4.0 termcolor==1.1.0 typing-extensions==3.7.4.3 urllib3==1.26.4 Werkzeug==1.0.1 wrapt==1.12.1
参照先には色々手順が書いてあるが、めんどくさいので パッチファイル用意した(足りない修正もあるし)。
以下の内容をssd_keras.patch ssd_keras_2.patchのファイル名で保存しておく。
参照先の情報によるpatch
ssd_keras.patch
diff --git a/ssd.py b/ssd.py
index c0049d8..656cd83 100644
--- a/ssd.py
+++ b/ssd.py
@@ -2,14 +2,15 @@
import keras.backend as K
from keras.layers import Activation
-from keras.layers import AtrousConvolution2D
+#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 merge
+#from keras.layers import merge
+from keras.layers.merge import concatenate
from keras.layers import Reshape
from keras.layers import ZeroPadding2D
from keras.models import Model
@@ -34,109 +35,115 @@ def SSD300(input_shape, num_classes=21):
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,
+ net['conv1_1'] = Convolution2D(64, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv1_1')(net['input'])
- net['conv1_2'] = Convolution2D(64, 3, 3,
+
+ net['conv1_2'] = Convolution2D(64, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv1_2')(net['conv1_1'])
- net['pool1'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',
+
+ net['pool1'] = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool1')(net['conv1_2'])
# Block 2
- net['conv2_1'] = Convolution2D(128, 3, 3,
+ net['conv2_1'] = Convolution2D(128, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv2_1')(net['pool1'])
- net['conv2_2'] = Convolution2D(128, 3, 3,
+
+ net['conv2_2'] = Convolution2D(128, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv2_2')(net['conv2_1'])
- net['pool2'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',
+ net['pool2'] = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool2')(net['conv2_2'])
+
# Block 3
- net['conv3_1'] = Convolution2D(256, 3, 3,
+ net['conv3_1'] = Convolution2D(256, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv3_1')(net['pool2'])
- net['conv3_2'] = Convolution2D(256, 3, 3,
+ net['conv3_2'] = Convolution2D(256, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv3_2')(net['conv3_1'])
- net['conv3_3'] = Convolution2D(256, 3, 3,
+ net['conv3_3'] = Convolution2D(256, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv3_3')(net['conv3_2'])
- net['pool3'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',
+ net['pool3'] = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool3')(net['conv3_3'])
# Block 4
- net['conv4_1'] = Convolution2D(512, 3, 3,
+ net['conv4_1'] = Convolution2D(512, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv4_1')(net['pool3'])
- net['conv4_2'] = Convolution2D(512, 3, 3,
+ net['conv4_2'] = Convolution2D(512, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv4_2')(net['conv4_1'])
- net['conv4_3'] = Convolution2D(512, 3, 3,
+ net['conv4_3'] = Convolution2D(512, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv4_3')(net['conv4_2'])
- net['pool4'] = MaxPooling2D((2, 2), strides=(2, 2), border_mode='same',
+ net['pool4'] = MaxPooling2D((2, 2), strides=(2, 2), padding='same',
name='pool4')(net['conv4_3'])
- # Block 5
- net['conv5_1'] = Convolution2D(512, 3, 3,
+# Block 5
+ net['conv5_1'] = Convolution2D(512, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv5_1')(net['pool4'])
- net['conv5_2'] = Convolution2D(512, 3, 3,
+ net['conv5_2'] = Convolution2D(512, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv5_2')(net['conv5_1'])
- net['conv5_3'] = Convolution2D(512, 3, 3,
+ net['conv5_3'] = Convolution2D(512, (3, 3),
activation='relu',
- border_mode='same',
+ padding='same',
name='conv5_3')(net['conv5_2'])
- net['pool5'] = MaxPooling2D((3, 3), strides=(1, 1), border_mode='same',
+ net['pool5'] = MaxPooling2D((3, 3), strides=(1, 1), padding='same',
name='pool5')(net['conv5_3'])
# FC6
- net['fc6'] = AtrousConvolution2D(1024, 3, 3, atrous_rate=(6, 6),
- activation='relu', border_mode='same',
+ net['fc6'] = Convolution2D(1024, (3, 3), dilation_rate=(6, 6),
+ activation='relu', padding='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'])
+ net['fc7'] = Convolution2D(1024, (1, 1), activation='relu',
+ padding='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',
+ net['conv6_1'] = Convolution2D(256, (1, 1), activation='relu',
+ padding='same',
name='conv6_1')(net['fc7'])
- net['conv6_2'] = Convolution2D(512, 3, 3, subsample=(2, 2),
- activation='relu', border_mode='same',
+
+
+ net['conv6_2'] = Convolution2D(512, (3, 3), strides=(2, 2),
+ activation='relu', padding='same',
name='conv6_2')(net['conv6_1'])
- # Block 7
- net['conv7_1'] = Convolution2D(128, 1, 1, activation='relu',
- border_mode='same',
+ # Block 7
+ net['conv7_1'] = Convolution2D(128, (1, 1), activation='relu',
+ padding='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',
+ net['conv7_2'] = Convolution2D(256, (3, 3), strides=(2, 2),
+ activation='relu', padding='valid',
name='conv7_2')(net['conv7_2'])
# Block 8
- net['conv8_1'] = Convolution2D(128, 1, 1, activation='relu',
- border_mode='same',
+ net['conv8_1'] = Convolution2D(128, (1, 1), activation='relu',
+ padding='same',
name='conv8_1')(net['conv7_2'])
- net['conv8_2'] = Convolution2D(256, 3, 3, subsample=(2, 2),
- activation='relu', border_mode='same',
+ net['conv8_2'] = Convolution2D(256, (3, 3), strides=(2, 2),
+ activation='relu', padding='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',
+ x = Convolution2D(num_priors * 4, (3, 3), padding='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')
@@ -144,7 +151,7 @@ def SSD300(input_shape, num_classes=21):
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',
+ x = Convolution2D(num_priors * num_classes, (3, 3), padding='same',
name=name)(net['conv4_3_norm'])
net['conv4_3_norm_mbox_conf'] = x
flatten = Flatten(name='conv4_3_norm_mbox_conf_flat')
@@ -155,16 +162,16 @@ def SSD300(input_shape, num_classes=21):
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',
+ net['fc7_mbox_loc'] = Convolution2D(num_priors * 4, (3, 3),
+ padding='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',
+ net['fc7_mbox_conf'] = Convolution2D(num_priors * num_classes, (3, 3),
+ padding='same',
name=name)(net['fc7'])
flatten = Flatten(name='fc7_mbox_conf_flat')
net['fc7_mbox_conf_flat'] = flatten(net['fc7_mbox_conf'])
@@ -174,7 +181,7 @@ def SSD300(input_shape, num_classes=21):
net['fc7_mbox_priorbox'] = priorbox(net['fc7'])
# Prediction from conv6_2
num_priors = 6
- x = Convolution2D(num_priors * 4, 3, 3, border_mode='same',
+ x = Convolution2D(num_priors * 4, (3, 3), padding='same',
name='conv6_2_mbox_loc')(net['conv6_2'])
net['conv6_2_mbox_loc'] = x
flatten = Flatten(name='conv6_2_mbox_loc_flat')
@@ -182,7 +189,7 @@ def SSD300(input_shape, num_classes=21):
name = 'conv6_2_mbox_conf'
if num_classes != 21:
name += '_{}'.format(num_classes)
- x = Convolution2D(num_priors * num_classes, 3, 3, border_mode='same',
+ x = Convolution2D(num_priors * num_classes, (3, 3), padding='same',
name=name)(net['conv6_2'])
net['conv6_2_mbox_conf'] = x
flatten = Flatten(name='conv6_2_mbox_conf_flat')
@@ -193,7 +200,7 @@ def SSD300(input_shape, num_classes=21):
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',
+ x = Convolution2D(num_priors * 4, (3, 3), padding='same',
name='conv7_2_mbox_loc')(net['conv7_2'])
net['conv7_2_mbox_loc'] = x
flatten = Flatten(name='conv7_2_mbox_loc_flat')
@@ -201,7 +208,7 @@ def SSD300(input_shape, num_classes=21):
name = 'conv7_2_mbox_conf'
if num_classes != 21:
name += '_{}'.format(num_classes)
- x = Convolution2D(num_priors * num_classes, 3, 3, border_mode='same',
+ x = Convolution2D(num_priors * num_classes, (3, 3), padding='same',
name=name)(net['conv7_2'])
net['conv7_2_mbox_conf'] = x
flatten = Flatten(name='conv7_2_mbox_conf_flat')
@@ -212,7 +219,7 @@ def SSD300(input_shape, num_classes=21):
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',
+ x = Convolution2D(num_priors * 4, (3, 3), padding='same',
name='conv8_2_mbox_loc')(net['conv8_2'])
net['conv8_2_mbox_loc'] = x
flatten = Flatten(name='conv8_2_mbox_loc_flat')
@@ -220,7 +227,7 @@ def SSD300(input_shape, num_classes=21):
name = 'conv8_2_mbox_conf'
if num_classes != 21:
name += '_{}'.format(num_classes)
- x = Convolution2D(num_priors * num_classes, 3, 3, border_mode='same',
+ x = Convolution2D(num_priors * num_classes, (3, 3), padding='same',
name=name)(net['conv8_2'])
net['conv8_2_mbox_conf'] = x
flatten = Flatten(name='conv8_2_mbox_conf_flat')
@@ -241,7 +248,7 @@ def SSD300(input_shape, num_classes=21):
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':
+ if K.image_data_format() == 'channels_last':
target_shape = (1, 1, 256)
else:
target_shape = (256, 1, 1)
@@ -249,42 +256,47 @@ def SSD300(input_shape, num_classes=21):
name='pool6_reshaped')(net['pool6'])
net['pool6_mbox_priorbox'] = priorbox(net['pool6_reshaped'])
# Gather all predictions
- net['mbox_loc'] = merge([net['conv4_3_norm_mbox_loc_flat'],
+ 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']],
- mode='concat', concat_axis=1, name='mbox_loc')
- net['mbox_conf'] = merge([net['conv4_3_norm_mbox_conf_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']],
- mode='concat', concat_axis=1, name='mbox_conf')
- net['mbox_priorbox'] = merge([net['conv4_3_norm_mbox_priorbox'],
+ 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']],
- mode='concat', concat_axis=1,
+ 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
+ num_boxes = net['mbox_loc'].int_shape()[-1] // 4
+ elif hasattr(net['mbox_loc'], 'get_shape'):
+ num_boxes = net['mbox_loc'].get_shape()[-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'] = merge([net['mbox_loc'],
+ net['predictions'] = concatenate([net['mbox_loc'],
net['mbox_conf'],
net['mbox_priorbox']],
- mode='concat', concat_axis=2,
+ axis=2,
+ #axis = 0,
name='predictions')
model = Model(net['input'], net['predictions'])
return model
diff --git a/ssd_layers.py b/ssd_layers.py
index 5e10478..41ee510 100644
--- a/ssd_layers.py
+++ b/ssd_layers.py
@@ -29,7 +29,8 @@ class Normalize(Layer):
Add possibility to have one scale for all features.
"""
def __init__(self, scale, **kwargs):
- if K.image_dim_ordering() == 'tf':
+ if K.image_data_format() == 'channels_last':
+
self.axis = 3
else:
self.axis = 1
@@ -41,7 +42,7 @@ class Normalize(Layer):
shape = (input_shape[self.axis],)
init_gamma = self.scale * np.ones(shape)
self.gamma = K.variable(init_gamma, name='{}_gamma'.format(self.name))
- self.trainable_weights = [self.gamma]
+ self._trainable_weights = [self.gamma]
def call(self, x, mask=None):
output = K.l2_normalize(x, self.axis)
@@ -81,7 +82,7 @@ class PriorBox(Layer):
"""
def __init__(self, img_size, min_size, max_size=None, aspect_ratios=None,
flip=True, variances=[0.1], clip=True, **kwargs):
- if K.image_dim_ordering() == 'tf':
+ if K.image_data_format() == 'channels_last':
self.waxis = 2
self.haxis = 1
else:
@@ -108,7 +109,7 @@ class PriorBox(Layer):
self.clip = True
super(PriorBox, self).__init__(**kwargs)
- def get_output_shape_for(self, input_shape):
+ def compute_output_shape(self, input_shape):
num_priors_ = len(self.aspect_ratios)
layer_width = input_shape[self.waxis]
layer_height = input_shape[self.haxis]
diff --git a/ssd_utils.py b/ssd_utils.py
index 0a9ffb1..4fc7a49 100644
--- a/ssd_utils.py
+++ b/ssd_utils.py
@@ -1,7 +1,8 @@
"""Some utils for SSD."""
import numpy as np
-import tensorflow as tf
+import tensorflow.compat.v1 as tf
+tf.disable_v2_behavior()
class BBoxUtility(object):
diff --git a/testing_utils/videotest.py b/testing_utils/videotest.py
index 0cbae3c..7b2ff4f 100644
--- a/testing_utils/videotest.py
+++ b/testing_utils/videotest.py
@@ -3,13 +3,14 @@
import cv2
import keras
from keras.applications.imagenet_utils import preprocess_input
-from keras.backend.tensorflow_backend import set_session
+# from keras.backend.tensorflow_backend import set_session
from keras.models import Model
from keras.preprocessing import image
import pickle
import numpy as np
from random import shuffle
-from scipy.misc import imread, imresize
+# from scipy.misc import imread, imresize
+from imageio import imread
from timeit import default_timer as timer
import sys
@@ -84,8 +85,8 @@ class VideoTest(object):
"trying to open a webcam, make sure you video_path is an integer!"))
# Compute aspect ratio of video
- vidw = vid.get(cv2.cv.CV_CAP_PROP_FRAME_WIDTH)
- vidh = vid.get(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT)
+ vidw = vid.get(cv2.CAP_PROP_FRAME_WIDTH)
+ vidh = vid.get(cv2.CAP_PROP_FRAME_HEIGHT)
vidar = vidw/vidh
# Skip frames until reaching start_frame
diff --git a/testing_utils/videotest_example.py b/testing_utils/videotest_example.py
index fb4d873..27ec148 100644
--- a/testing_utils/videotest_example.py
+++ b/testing_utils/videotest_example.py
@@ -1,3 +1,7 @@
+import os
+# GPU不使用を設定
+os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
+
import keras
import pickle
from videotest import VideoTest
@@ -9,16 +13,38 @@ from ssd import SSD300 as SSD
input_shape = (300,300,3)
# Change this if you run with other classes than VOC
-class_names = ["background", "aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"];
+class_names = [
+ "background",
+ "aeroplane",
+ "bicycle",
+ "bird",
+ "boat",
+ "bottle",
+ "bus",
+ "car",
+ "cat",
+ "chair",
+ "cow",
+ "diningtable",
+ "dog",
+ "horse",
+ "motorbike",
+ "person",
+ "pottedplant",
+ "sheep",
+ "sofa",
+ "train",
+ "tvmonitor"
+ ];
NUM_CLASSES = len(class_names)
model = SSD(input_shape, num_classes=NUM_CLASSES)
# Change this path if you want to use your own trained weights
-model.load_weights('../weights_SSD300.hdf5')
+model.load_weights('../../weights_SSD300.hdf5')
vid_test = VideoTest(class_names, model, input_shape)
# To test on webcam 0, remove the parameter (or change it to another number
# to test on that webcam)
-vid_test.run('path/to/your/video.mkv')
+vid_test.run('../../testvideo3.mp4')
keras修正対応
ssd_keras_2.patch
diff -ur '--exclude=.git' '--exclude=__pycache__' ssd_keras/ssd.py ssd_keras.tmp/ssd.py
--- ssd_keras/ssd.py 2021-05-21 07:13:41.880000000 +0900
+++ ssd_keras.tmp/ssd.py 2021-05-21 07:06:44.970000000 +0900
@@ -1,19 +1,17 @@
"""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 merge
-from keras.layers.merge import concatenate
-from keras.layers import Reshape
-from keras.layers import ZeroPadding2D
-from keras.models import Model
+import tensorflow.keras.backend as K
+from tensorflow.keras.layers import Activation
+from tensorflow.keras.layers import Convolution2D
+from tensorflow.keras.layers import Dense
+from tensorflow.keras.layers import Flatten
+from tensorflow.keras.layers import GlobalAveragePooling2D
+from tensorflow.keras.layers import Input
+from tensorflow.keras.layers import MaxPooling2D
+from tensorflow.keras.layers import concatenate
+from tensorflow.keras.layers import Reshape
+from tensorflow.keras.layers import ZeroPadding2D
+from tensorflow.keras.models import Model
from ssd_layers import Normalize
from ssd_layers import PriorBox
diff -ur '--exclude=.git' '--exclude=__pycache__' ssd_keras/ssd_layers.py ssd_keras.tmp/ssd_layers.py
--- ssd_keras/ssd_layers.py 2021-05-21 07:13:41.880000000 +0900
+++ ssd_keras.tmp/ssd_layers.py 2021-05-21 06:31:11.780000000 +0900
@@ -1,8 +1,8 @@
"""Some special pupropse layers for SSD."""
-import keras.backend as K
-from keras.engine.topology import InputSpec
-from keras.engine.topology import Layer
+import tensorflow.keras.backend as K
+from tensorflow.python.keras.engine.input_spec import InputSpec
+from tensorflow.python.keras.engine.base_layer import Layer
import numpy as np
import tensorflow as tf
diff -ur '--exclude=.git' '--exclude=__pycache__' ssd_keras/testing_utils/videotest.py ssd_keras.tmp/testing_utils/videotest.py
--- ssd_keras/testing_utils/videotest.py 2021-05-21 07:13:41.880000000 +0900
+++ ssd_keras.tmp/testing_utils/videotest.py 2021-05-21 07:08:24.680000000 +0900
@@ -1,15 +1,13 @@
""" A class for testing a SSD model on a video file or webcam """
import cv2
-import keras
-from keras.applications.imagenet_utils import preprocess_input
-# from keras.backend.tensorflow_backend import set_session
-from keras.models import Model
-from keras.preprocessing import image
+import tensorflow.keras as keras
+from tensorflow.keras.applications.imagenet_utils import preprocess_input
+from tensorflow.keras.models import Model
+from tensorflow.keras.preprocessing import image
import pickle
import numpy as np
from random import shuffle
-# from scipy.misc import imread, imresize
from imageio import imread
from timeit import default_timer as timer
@@ -179,6 +177,7 @@
cv2.putText(to_draw, fps, (3,10), cv2.FONT_HERSHEY_SIMPLEX, 0.35, (0,0,0), 1)
cv2.imshow("SSD result", to_draw)
- cv2.waitKey(10)
-
-
+ key = cv2.waitKey(1)
+ if key == 27:
+ # ESCキー
+ break
diff -ur '--exclude=.git' '--exclude=__pycache__' ssd_keras/testing_utils/videotest_example.py ssd_keras.tmp/testing_utils/videotest_example.py
--- ssd_keras/testing_utils/videotest_example.py 2021-05-21 07:13:41.880000000 +0900
+++ ssd_keras.tmp/testing_utils/videotest_example.py 2021-05-21 07:04:24.320000000 +0900
@@ -2,7 +2,7 @@
# GPU不使用を設定
os.environ["CUDA_VISIBLE_DEVICES"] = "-1"
-import keras
+import tensorflow.keras as keras
import pickle
from videotest import VideoTest
@@ -47,4 +47,4 @@
# To test on webcam 0, remove the parameter (or change it to another number
# to test on that webcam)
-vid_test.run('../../testvideo3.mp4')
+vid_test.run('../../images/testvideo3.mp4')
参照先の記載にしたがって、重み学習済み重みファイルをダウンロードする。
wgetでは取得できなかったので、ブラウザで↓ここ から weights_SSD300.hdf5 をダウンロード。
https://mega.nz/folder/7RowVLCL#q3cEVRK9jyOSB9el3SssIA
で、ZIPファイルをカレントディレクトリに解凍しておく。
githubからソースをダウンロードする。
このままではエラーになるので、先ほど作成したパッチファイルssd_keras.patchでパッチをあてる。
git clone https://github.com/rykov8/ssd_keras.git
cd ssd_keras
patch -p1 < ../ssd_keras.patch
# Keras修正対応パッチ
patch -p1 < ../ssd_keras_2.patch
実行する。
cd testing_utils/
python videotest_example.py
おー。
しかし遅い….
Kerasは遅いみたい。。。orz…