Python の URL: http://www.python.org/
インストール手順の詳細は： 別ページで説明している．
Windows では，Python 3.8 や 3.6 の共存は簡単．
コマンドプロンプトを管理者として実行し，次のコマンドを実行．
python -m pip install -U pip setuptools
python -m pip uninstall -y tensorflow tensorflow-cpu tensorflow-gpu tensorflow_datasets tensorflow-hub keras python -m pip install -U tensorflow-gpu==1.15.3 tensorflow_datasets python -m pip install -U keras==2.2.4 opencv-python matplotlib numpy imageio tqdm beautifulsoup4 scikit-learn python -m pip install -U jupyterlab jupyter jupyter-console jupytext spyder
画像ファイル fruits.jpg, home.jpg のダウンロード手順は，別ページで説明している．
https://github.com/opencv/opencv/tree/master/samples/data で公開されている fruits.jpg, home.jpg を使用する（謝辞：画像の作者に感謝します）
SSD の原論文 (https://arxiv.org/abs/1512.02325) では VGG-16 を使用．
https://github.com/pierluigiferrari/ssd_keras では SSD のプログラム、 Pascal VOC2007, Pascal VOC2012, COCO, ILSVRC を用いての学習済みニューラルネットワークの重みデータが配布されている
Pascal VOC2007, Pascal COC2012 は、画像約10000個と、それぞれ20種類のラベルのデータである
mkdir c:\pytools cd c:\pytools rmdir /s /q ssd_keras
cd c:\pytools git clone https://github.com/pierluigiferrari/ssd_keras
ファイル名は、 VGG_coco_SSD_300x300_iter_400000.h5
ファイル名は、 VGG_VOC0712_SSD_300x300_iter_120000.h5
このWebページの下の方にある
以上で準備ができた
謝辞：https://github.com/pierluigiferrari/ssd_keras/blob/master/ssd300_inference.ipynb に記載のプログラムを使用している
Python のプログラム作成には，開発環境や Python コンソール（Jupyter Qt Console，spyder，PyCharm，PyScripter など）が便利．
COCO を用いての学習済みニューラルネットワークの重みデータを使う。
cd C:/pytools/ssd_keras
from keras import backend as K from keras.models import load_model from tensorflow.keras.preprocessing import image from keras.optimizers import Adam from imageio import imread import numpy as np from matplotlib import pyplot as plt import sys from models.keras_ssd300 import ssd_300 from keras_loss_function.keras_ssd_loss import SSDLoss from keras_layers.keras_layer_AnchorBoxes import AnchorBoxes from keras_layers.keras_layer_DecodeDetections import DecodeDetections from keras_layers.keras_layer_DecodeDetectionsFast import DecodeDetectionsFast from keras_layers.keras_layer_L2Normalization import L2Normalization from ssd_encoder_decoder.ssd_output_decoder import decode_detections, decode_detections_fast from data_generator.object_detection_2d_data_generator import DataGenerator from data_generator.object_detection_2d_photometric_ops import ConvertTo3Channels from data_generator.object_detection_2d_geometric_ops import Resize from data_generator.object_detection_2d_misc_utils import apply_inverse_transforms # Set the image size. img_height = 300 img_width = 300
![[image]](../../info/logo_png.png){kind=logo}
# 1: Build the Keras model
K.clear_session()
model = ssd_300(image_size=(img_height, img_width, 3),
n_classes=80,
mode='inference',
l2_regularization=0.0005,
scales=[0.07, 0.15, 0.33, 0.51, 0.69, 0.87, 1.05], # The scales for Pascal VOC are [0.1, 0.2, 0.37, 0.54, 0.71, 0.88, 1.05]
aspect_ratios_per_layer=[[1.0, 2.0, 0.5],
[1.0, 2.0, 0.5, 3.0, 1.0/3.0],
[1.0, 2.0, 0.5, 3.0, 1.0/3.0],
[1.0, 2.0, 0.5, 3.0, 1.0/3.0],
[1.0, 2.0, 0.5],
[1.0, 2.0, 0.5]],
two_boxes_for_ar1=True,
steps=[8, 16, 32, 64, 100, 300],
offsets=[0.5, 0.5, 0.5, 0.5, 0.5, 0.5],
clip_boxes=False,
variances=[0.1, 0.1, 0.2, 0.2],
normalize_coords=True,
subtract_mean=[123, 117, 104],
swap_channels=[2, 1, 0],
confidence_thresh=0.01,
iou_threshold=0.45,
top_k=200,
nms_max_output_size=400)
# 2: Load the trained weights into the model.
# TODO: Set the path of the trained weights.
weights_path = 'C:/pytools/ssd_keras/VGG_coco_SSD_300x300_iter_400000.h5'
model.load_weights(weights_path, by_name=True)
# 3: Compile the model so that Keras won't complain the next time you load it.
adam = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)
ssd_loss = SSDLoss(neg_pos_ratio=3, alpha=1.0)
m.compile(optimizer=adam, loss=ssd_loss.compute_loss)
画像ファイル名は、「img_path = 'C:/image/home.jpg'」の行で設定
colors = plt.cm.hsv(np.linspace(0, 1, 81)).tolist()
orig_images = [] # Store the images here.
input_images = [] # Store resized versions of the images here.
# We'll only load one image in this example.
img_path = 'C:/image/home.jpg'
orig_images.append(imread(img_path))
img = image.load_img(img_path, target_size=(img_height, img_width))
img = image.img_to_array(img)
input_images.append(img)
input_images = np.array(input_images)
y_pred = model.predict(input_images)
confidence_threshold = 0.075
y_pred_thresh = [y_pred[k][y_pred[k,:,1] > confidence_threshold] for k in range(y_pred.shape[0])]
plt.figure(figsize=(20,12))
plt.imshow(orig_images[0])
current_axis = plt.gca()
for box in y_pred_thresh[0]:
# Transform the predicted bounding boxes for the 300x300 image to the original image dimensions.
xmin = box[2] * orig_images[0].shape[1] / img_width
ymin = box[3] * orig_images[0].shape[0] / img_height
xmax = box[4] * orig_images[0].shape[1] / img_width
ymax = box[5] * orig_images[0].shape[0] / img_height
color = colors[int(box[0])]
label = '{}: {:.2f}'.format(box[0], box[1])
current_axis.add_patch(plt.Rectangle((xmin, ymin), xmax-xmin, ymax-ymin, color=color, fill=False, linewidth=2))
current_axis.text(xmin, ymin, label, size='x-large', color='white', bbox={'facecolor':color, 'alpha':1.0})
結果の確認
プログラムでは「img_path = 'C:/image/fruits.jpg'」のように変えて、実行（他の部分は変えない）
colors = plt.cm.hsv(np.linspace(0, 1, 81)).tolist()
orig_images = [] # Store the images here.
input_images = [] # Store resized versions of the images here.
# We'll only load one image in this example.
img_path = 'C:/image/fruits.jpg'
orig_images.append(imread(img_path))
img = image.load_img(img_path, target_size=(img_height, img_width))
img = img.ravel()
input_images = np.array(img)
y_pred = model.predict(input_images)
confidence_threshold = 0.1
y_pred_thresh = [y_pred[k][y_pred[k,:,1] > confidence_threshold] for k in range(y_pred.shape[0])]
plt.figure(figsize=(20,12))
plt.imshow(orig_images[0])
current_axis = plt.gca()
for box in y_pred_thresh[0]:
# Transform the predicted bounding boxes for the 300x300 image to the original image dimensions.
xmin = box[2] * orig_images[0].shape[1] / img_width
ymin = box[3] * orig_images[0].shape[0] / img_height
xmax = box[4] * orig_images[0].shape[1] / img_width
ymax = box[5] * orig_images[0].shape[0] / img_height
color = colors[int(box[0])]
label = '{}: {:.2f}'.format(box[0], box[1])
current_axis.add_patch(plt.Rectangle((xmin, ymin), xmax-xmin, ymax-ymin, color=color, fill=False, linewidth=2))
current_axis.text(xmin, ymin, label, size='x-large', color='white', bbox={'facecolor':color, 'alpha':1.0})
プログラムでは「img_path = 'C:/image/1.png'」のように変えて、実行（他の部分は変えない）
colors = plt.cm.hsv(np.linspace(0, 1, 81)).tolist()
orig_images = [] # Store the images here.
input_images = [] # Store resized versions of the images here.
# We'll only load one image in this example.
img_path = 'C:/image/1.png'
orig_images.append(imread(img_path))
img = image.load_img(img_path, target_size=(img_height, img_width))
img = image.img_to_array(img)
input_images.append(img)
input_images = np.array(input_images)
y_pred = model.predict(input_images)
confidence_threshold = 0.1
y_pred_thresh = [y_pred[k][y_pred[k,:,1] > confidence_threshold] for k in range(y_pred.shape[0])]
plt.figure(figsize=(20,12))
plt.imshow(orig_images[0])
current_axis = plt.gca()
for box in y_pred_thresh[0]:
# Transform the predicted bounding boxes for the 300x300 image to the original image dimensions.
xmin = box[2] * orig_images[0].shape[1] / img_width
ymin = box[3] * orig_images[0].shape[0] / img_height
xmax = box[4] * orig_images[0].shape[1] / img_width
ymax = box[5] * orig_images[0].shape[0] / img_height
color = colors[int(box[0])]
label = '{}: {:.2f}'.format(box[0], box[1])
current_axis.add_patch(plt.Rectangle((xmin, ymin), xmax-xmin, ymax-ymin, color=color, fill=False, linewidth=2))
current_axis.text(xmin, ymin, label, size='x-large', color='white', bbox={'facecolor':color, 'alpha':1.0})
謝辞：https://github.com/pierluigiferrari/ssd_keras/blob/master/ssd300_inference.ipynb に記載のプログラムを使用している
Python のプログラム作成には，開発環境や Python コンソール（Jupyter Qt Console，spyder，PyCharm，PyScripter など）が便利．
cd C:/pytools/ssd_keras
from keras import backend as K from keras.models import load_model from tensorflow.keras.preprocessing import image from keras.optimizers import Adam from imageio import imread import numpy as np from matplotlib import pyplot as plt from models.keras_ssd300 import ssd_300 from keras_loss_function.keras_ssd_loss import SSDLoss from keras_layers.keras_layer_AnchorBoxes import AnchorBoxes from keras_layers.keras_layer_DecodeDetections import DecodeDetections from keras_layers.keras_layer_DecodeDetectionsFast import DecodeDetectionsFast from keras_layers.keras_layer_L2Normalization import L2Normalization from ssd_encoder_decoder.ssd_output_decoder import decode_detections, decode_detections_fast from data_generator.object_detection_2d_data_generator import DataGenerator from data_generator.object_detection_2d_photometric_ops import ConvertTo3Channels from data_generator.object_detection_2d_geometric_ops import Resize from data_generator.object_detection_2d_misc_utils import apply_inverse_transforms # Set the image size. img_height = 300 img_width = 300
# 1: Build the Keras model
K.clear_session() # Clear previous models from memory.
model = ssd_300(image_size=(img_height, img_width, 3),
n_classes=20,
mode='inference',
l2_regularization=0.0005,
scales=[0.1, 0.2, 0.37, 0.54, 0.71, 0.88, 1.05], # The scales for MS COCO are [0.07, 0.15, 0.33, 0.51, 0.69, 0.87, 1.05]
aspect_ratios_per_layer=[[1.0, 2.0, 0.5],
[1.0, 2.0, 0.5, 3.0, 1.0/3.0],
[1.0, 2.0, 0.5, 3.0, 1.0/3.0],
[1.0, 2.0, 0.5, 3.0, 1.0/3.0],
[1.0, 2.0, 0.5],
[1.0, 2.0, 0.5]],
two_boxes_for_ar1=True,
steps=[8, 16, 32, 64, 100, 300],
offsets=[0.5, 0.5, 0.5, 0.5, 0.5, 0.5],
clip_boxes=False,
variances=[0.1, 0.1, 0.2, 0.2],
normalize_coords=True,
subtract_mean=[123, 117, 104],
swap_channels=[2, 1, 0],
confidence_thresh=0.5,
iou_threshold=0.45,
top_k=200,
nms_max_output_size=400)
# 2: Load the trained weights into the model.
# TODO: Set the path of the trained weights.
weights_path = 'C:/pytools/ssd_keras/VGG_VOC0712_SSD_300x300_iter_120000.h5'
model.load_weights(weights_path, by_name=True)
# 3: Compile the model so that Keras won't complain the next time you load it.
adam = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)
ssd_loss = SSDLoss(neg_pos_ratio=3, alpha=1.0)
m.compile(optimizer=adam, loss=ssd_loss.compute_loss)
colors = plt.cm.hsv(np.linspace(0, 1, 21)).tolist()
classes = ['background',
'aeroplane', 'bicycle', 'bird', 'boat',
'bottle', 'bus', 'car', 'cat',
'chair', 'cow', 'diningtable', 'dog',
'horse', 'motorbike', 'person', 'pottedplant',
'sheep', 'sofa', 'train', 'tvmonitor']
orig_images = [] # Store the images here.
input_images = [] # Store resized versions of the images here.
# We'll only load one image in this example.
img_path = 'C:/image/home.jpg'
orig_images.append(imread(img_path))
img = image.load_img(img_path, target_size=(img_height, img_width))
img = image.img_to_array(img)
input_images.append(img)
input_images = np.array(input_images)
y_pred = model.predict(input_images)
confidence_threshold = 0.5
y_pred_thresh = [y_pred[k][y_pred[k,:,1] > confidence_threshold] for k in range(y_pred.shape[0])]
plt.figure(figsize=(20,12))
plt.imshow(orig_images[0])
current_axis = plt.gca()
for box in y_pred_thresh[0]:
# Transform the predicted bounding boxes for the 300x300 image to the original image dimensions.
xmin = box[2] * orig_images[0].shape[1] / img_width
ymin = box[3] * orig_images[0].shape[0] / img_height
xmax = box[4] * orig_images[0].shape[1] / img_width
ymax = box[5] * orig_images[0].shape[0] / img_height
color = colors[int(box[0])]
label = '{}: {:.2f}'.format(classes[int(box[0])], box[1])
current_axis.add_patch(plt.Rectangle((xmin, ymin), xmax-xmin, ymax-ymin, color=color, fill=False, linewidth=2))
current_axis.text(xmin, ymin, label, size='x-large', color='white', bbox={'facecolor':color, 'alpha':1.0})
結果を確認
本サイトは金子邦彦研究室のWebページである．
資料等の公開では，原則，「クリエイティブコモンズ BY NC SA」として公開するようにしている． PDFファイル，パワーポイントファイルなどには， 「クリエイティブコモンズ BY NC SA」を明記するとともに，ロゴを記載するようにしている（作業が間に合っていない分もあるのでご容赦ください）．
公開している資料をご利用になる場合の，再配布の条件，剽窃の防止などについて，別ページで説明している． 再配布や資料改変の際には，そのページをご確認ください．
サイトマップは，サイトマップのページをご覧下さい． 本サイト内の検索は，サイト内検索のページをご利用下さい．
問い合わせ先： 金子邦彦（かねこ くにひこ） ![[image]](../../info/kanekomail.png){kind=small}