顔認識のデモ#

概要#

スクリプト face_detection.py は、input_images_dir ディレクトリーに保存されているすべての画像に対して顔検出推論要求を実行します。

スクリプトは入力画像のサイズを調整し、要求内のバッチサイズを変更できます。OpenVINO モデルサーバーの動的形状の機能を使用する方法と、サーバーからの出力を処理する方法を示します。

この例は、モデル face-detection-retail-0004 に依存しています。

リポジトリーのクローンを作成し、face_detection ディレクトリーに入ります

git clone https://github.com/openvinotoolkit/model_server.git 
cd model_server/demos/face_detection/python

環境の準備:

pip install -r ../../common/python/requirements.txt

python face_detection.py --help 
usage: face_detection.py [-h] [--input_images_dir INPUT_IMAGES_DIR] 
                         [--output_dir OUTPUT_DIR] [--batch_size BATCH_SIZE] 
                         [--width WIDTH] [--height HEIGHT] 
                         [--grpc_address GRPC_ADDRESS] [--grpc_port GRPC_PORT] 
                         [--model_name MODEL_NAME] [--tls] 
                         [--server_cert SERVER_CERT] 
                         [--client_cert CLIENT_CERT] [--client_key CLIENT_KEY] 

Demo for face detection requests via TFS gRPC API analyses input images and saves images with bounding boxes drawn around detected faces.It relies on face_detection model... 

optional arguments: 
  -h, --help            show this help message and exit 
  --input_images_dir INPUT_IMAGES_DIR 
                        Directory with input images 
  --output_dir OUTPUT_DIR
                        Directory for storing images with detection results 
  --batch_size BATCH_SIZE 
                        How many images should be grouped in one batch 
  --width WIDTH         How the input image width should be resized in pixels 
  --height HEIGHT       How the input image width should be resized in pixels 
  --grpc_address GRPC_ADDRESS 
                        Specify url to grpc service. Default:localhost 
  --grpc_port GRPC_PORT 
                        Specify port to grpc service. default: 9000 
  --model_name MODEL_NAME 
                        Specify the model name 
  --tls                 use TLS communication with gRPC endpoint 
  --server_cert SERVER_CERT 
                        Path to server certificate 
  --client_cert CLIENT_CERT 
                        Path to client certificate 
  --client_key CLIENT_KEY 
                        Path to client key

使用例#

OVMS サービスをローカルで開始します:

mkdir -p model/1 
wget -P model/1 https://storage.openvinotoolkit.org/repositories/open_model_zoo/2022.1/models_bin/2/face-detection-retail-0004/FP32/face-detection-retail-0004.bin 
wget -P model/1 https://storage.openvinotoolkit.org/repositories/open_model_zoo/2022.1/models_bin/2/face-detection-retail-0004/FP32/face-detection-retail-0004.xml 
docker run --rm -d -u $(id -u):$(id -g) -v `pwd`/model:/models -p 9000:9000 openvino/model_server:latest --model_path /models --model_name face-detection --port 9000 --shape auto

クライアントの実行:

mkdir results 

python face_detection.py --batch_size 1 --width 300 --height 300 --grpc_port 9000 

['people3.jpeg', 'people1.jpeg', 'people4.jpeg', 'people2.jpeg'] 
Start processing 4 iterations with batch size 1 

Request shape (1, 3, 300, 300) 
Response shape (1, 1, 200, 7) 
image in batch item 0 , output shape (3, 300, 300) 
detection 0 [[[0.1.0.9999211 0.45312274 0.1578588 0.5795723 0.4229777 ]]] 
x_min 135 
y_min 47 
x_max 173 
y_max 126 
detection 1 [[[0.1.0.99940693 0.22251067 0.22310418 0.34818068 0.50511265]]] 
x_min 66 
y_min 66 
x_max 104 
y_max 151 
detection 2 [[[0.1.0.98198867 0.61662686 0.08173735 0.740975 0.40580124]]] 
x_min 184 
y_min 24 
x_max 222 y_max 121 
saving result to results/1_0.jpg 
Iteration 1; Processing time: 28.69 ms; speed 34.85 fps 

Request shape (1, 3, 300, 300) 
Response shape (1, 1, 200, 7) 
image in batch item 0 , output shape (3, 300, 300) 
detection 0 [[[0.1.0.9999999 0.22627862 0.35042182 0.27032945 0.43312052]]] 
x_min 67 
y_min 105 
x_max 81 
y_max 129 
detection 1 [[[0.1.0.9999999 0.7980574 0.35572374 0.8422255 0.42749226]]] 
x_min 239 
y_min 106 
x_max 252 
y_max 128 
detection 2 [[[0.1.0.9999927 0.4413453 0.29417545 0.48191014 0.37180012]]] 
x_min 132 
y_min 88 
x_max 144 
y_max 111 
detection 3 [[[0.1.0.99964225 0.55356365 0.30400735 0.59468836 0.38264883]]] 
x_min 166 
y_min 91 
x_max 178 
y_max 114 
detection 4 [[[0.1.0.9993523 0.32912934 0.38222942 0.36873418 0.44978413]]] 
x_min 98 
y_min 114 
x_max 110 
y_max 134 
detection 5 [[[0.1.0.9992501 0.33522347 0.6249954 0.38323137 0.7104612 ]]] 
x_min 100 
y_min 187 
x_max 114 
y_max 213 
detection 6 [[[0.1.0.9976745 0.6488881 0.5992611 0.6988456 0.6907843]]] 
x_min 194 
y_min 179 
x_max 209 
y_max 207 
detection 7 [[[0.1.0.9962077 0.5180316 0.5640176 0.5703776 0.6516389]]] 
x_min 155 
y_min 169 
x_max 171 
y_max 195 
detection 8 [[[0.1.0.722986 0.6746904 0.3287916 0.7198625 0.4061382]]] 
x_min 202 
y_min 98 
x_max 215 
y_max 121 
detection 9 [[[0.1.0.566281 0.13994813 0.36546633 0.18363091 0.44829145]]] 
x_min 41 
y_min 109 
x_max 55 
y_max 134 
saving result to results/2_0.jpg 
Iteration 2; Processing time: 28.07 ms; speed 35.62 fps 

Request shape (1, 3, 300, 300) 
Response shape (1, 1, 200, 7) 
image in batch item 0 , output shape (3, 300, 300) 
detection 0 [[[0.1.0.99953306 0.1962457 0.08444536 0.33928737 0.43830967]]] 
x_min 58 
y_min 25 
x_max 101 
y_max 131 
detection 1 [[[0.1.0.9994074 0.7784856 0.12126188 0.9029822 0.38237268]]] 
x_min 233 
y_min 36 
x_max 270 
y_max 114 
saving result to results/3_0.jpg 
Iteration 3; Processing time: 31.37 ms; speed 31.88 fps 

Request shape (1, 3, 300, 300) 
Response shape (1, 1, 200, 7) 
image in batch item 0 , output shape (3, 300, 300) 
detection 0 [[[0.1.0.99999976 0.46376142 0.10504608 0.62492824 0.4838 ]]] 
x_min 139 
y_min 31 
x_max 187 
y_max 145 
detection 1 [[[0.1.0.9998708 0.12117466 0.53293884 0.2551119 0.753091 ]]] 
x_min 36 
y_min 159 
x_max 76 
y_max 225 
saving result to results/4_0.jpg 
Iteration 4; Processing time: 26.14 ms; speed 38.26 fps 

processing time for all iterations 
average time: 28.25 ms; average speed: 35.40 fps median time: 28.00 ms; 
median speed: 35.71 fps 
max time: 31.00 ms; min speed: 32.26 fps 
min time: 26.00 ms; max speed: 38.46 fps 
time percentile 90: 30.10 ms; speed percentile 90: 33.22 fps 
time percentile 50: 28.00 ms; speed percentile 50: 35.71 fps 
time standard deviation: 1.79 
time variance: 3.19

python face_detection.py --batch_size 4 --width 600 --height 400 --input_images_dir ../../common/static/images/people --output_dir results --grpc_port 9000 

['people3.jpeg', 'people1.jpeg', 'people4.jpeg', 'people2.jpeg'] 
Start processing 1 iterations with batch size 4 

Request shape (4, 3, 400, 600) 
Response shape (1, 1, 800, 7) 
image in batch item 0 , output shape (3, 400, 600) 
detection 0 [[[0.1.1.0.60533124 0.06616803 0.7431837 0.39988554]]] 
x_min 363 
y_min 26 
x_max 445 
y_max 159 
detection 1 [[[0.1.1.0.4518694 0.14568813 0.5847517 0.41192806]]] 
x_min 271 
y_min 58 
x_max 350 
y_max 164 
detection 2 [[[0.1.1.0.21670896 0.21642502 0.34729022 0.4911797 ]]] 
x_min 130 
y_min 86 x_max 208 
y_max 196 
saving result to results/1_0.jpg 
image in batch item 1 , output shape (3, 400, 600) 
detection 47 [[[1.1.1.0.55241716 0.30246916 0.59122956 0.3917096 ]]] 
x_min 331 
y_min 120 
x_max 354 
y_max 156 
detection 48 [[[1.1.0.9999999 0.33650026 0.6238751 0.38452044 0.71090096]]] 
x_min 201 
y_min 249 
x_max 230 
y_max 284 
detection 49 [[[1.1.0.99999917 0.22734313 0.34604052 0.2695081 0.44246814]]] 
x_min 136 
y_min 138 
x_max 161 
y_max 176 
detection 50 [[[1.1.0.99999905 0.4421842 0.29369104 0.4823516 0.3778398 ]]] 
x_min 265 
y_min 117 
x_max 289 
y_max 151 
detection 51 [[[1.1.0.9999931 0.51710206 0.56031024 0.5672712 0.6503216 ]]] 
x_min 310 
y_min 224 
x_max 340 
y_max 260 
detection 52 [[[1.1.0.99996173 0.79854035 0.35696298 0.8411697 0.42877442]]] 
x_min 479 
y_min 142 
x_max 504 
y_max 171 
detection 53 [[[1.1.0.99993885 0.65063936 0.59173524 0.6996588 0.6923708 ]]] 
x_min 390 
y_min 236 
x_max 419 
y_max 276 
detection 54 [[[1.1.0.99990237 0.67253107 0.3239646 0.7202968 0.41418064]]] 
x_min 403 
y_min 129 
x_max 432 
y_max 165 
detection 55 [[[1.1.0.99935323 0.13880333 0.36219725 0.18785533 0.45266667]]] 
x_min 83 
y_min 144 
x_max 112 
y_max 181 
detection 56 [[[1.1.0.9158124 0.3301806 0.37731093 0.3695439 0.44581354]]] 
x_min 198 
y_min 150 
x_max 221 
y_max 178 
saving result to results/1_1.jpg 
image in batch item 2 , output shape (3, 400, 600) 
detection 196 [[[2.1.1.0.1979972 0.09549272 0.33667618 0.43583226]]] 
x_min 118 
y_min 38 
x_max 202 
y_max 174 
detection 197 [[[2.1.0.9994671 0.78329325 0.11969317 0.90646887 0.38130802]]] 
x_min 469 
y_min 47 
x_max 543 
y_max 152 
saving result to results/1_2.jpg 
image in batch item 3 , output shape (3, 400, 600) 
detection 299 [[[3.1.1.0.45842466 0.11277443 0.6309127 0.48450887]]] 
x_min 275 
y_min 45 
x_max 378 
y_max 193 
detection 300 [[[3.1.0.999944 0.119257 0.5222848 0.2530746 0.75714105]]] 
x_min 71 
y_min 208 
x_max 151 
y_max 302 
saving result to results/1_3.jpg 
Iteration 1; Processing time: 325.79 ms; speed 12.28 fps 

processing time for all iterations 
average time: 325.00 ms; average speed: 12.31 fps 
median time: 325.00 ms; median speed: 12.31 fps 
max time: 325.00 ms; min speed: 12.31 fps 
min time: 325.00 ms; max speed: 12.31 fps 
time percentile 90: 325.00 ms; speed percentile 90: 12.31 fps 
time percentile 50: 325.00 ms; speed percentile 50: 12.31 fps 
time standard deviation: 0.00 
time variance: 0.00

スクリプトは、output_dir ディレクトリーに保存された画像上の推論結果を視覚化します。保存された画像には次の命名規則があります:

[iteration]_[image_in_batch].jpeg