Tengine Lite 的 examples 将提供简单的、好玩的 demo。
- 分类任务
- 人脸关键点检测任务
- ssd 目标检测任务
- retinaface 人脸检测任务
- yolact 实例分割任务
- yolov3 目标检测任务
- yolov4-tiny目标检测任务
- openpose人体姿态识别任务
- crnn汉字识别任务
分类任务 - tm_classification.c
Tengine Lite 兼容 Tengine 原有的 C API 供用户使用，这里我们使用 C API 展示如何运行 tm_classification 例程运行 MobileNet v1 分类网络模型，实现指定图片分类的功能。让你快速上手 Tengine Lite C API。这里，我们使用在这个撸猫时代行业从业者大爱的 tiger cat 作为测试图片。
模型在此处可以找到：Tengine model zoo 兼容原有 Tengine 的模型示例仓库（密码：hhgc）。
build.sh 编译脚本默认配置已实现自动编译 examples 中的 demo 程序，以 x86 平台为例，demo 存放在 ./build/install/bin/ 目录下。
bug1989@DESKTOP-SGN0H2A:/mnt/d/ubuntu/gitlab/build-linux$ tree install
install
├── bin
│ ├── cpp_tm_classification
│ ├── cpp_tm_mobilenet_ssd
│ ├── tm_benchmark
│ ├── tm_classification
│ ├── tm_classification_fp16
│ ├── tm_classification_uint8
│ ├── tm_classification_vulkan
│ ├── tm_crnn
│ ├── tm_landmark
│ ├── tm_landmark_uint8
│ ├── tm_mobilefacenet
│ ├── tm_mobilenet_ssd
│ ├── tm_mobilenet_ssd_acl
│ ├── tm_mobilenet_ssd_uint8
│ ├── tm_openpose
│ ├── tm_retinaface
│ ├── tm_yolact
│ ├── tm_yolov3_tiny
│ ├── tm_yolov3_uint8
│ ├── tm_yolov4
│ └── tm_yolov4_tiny
├── include
│ └── tengine_c_api.h
└── lib
└── libtengine-lite.so将测试图片和模型文件放在 Tengine-Lite 根目录下，运行：
export LD_LIBRARY_PATH=./build/install/lib
./build/install/bin/tm_classification -m models/mobilenet.tmfile -i images/cat.jpg -g 224,224 -s 0.017,0.017,0.017 -w 104.007,116.669,122.679结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
model file : ./temp/models/mobilenet.tmfile
image file : ./temp/images/cat.jpg
img_h, img_w, scale[3], mean[3] : 224 224 , 0.017 0.017 0.017, 104.0 116.7 122.7
Repeat 1 times, thread 1, avg time 656.76 ms, max_time 656.76 ms, min_time 656.76 ms
--------------------------------------
8.574148, 282
7.880116, 277
7.812579, 278
7.286453, 263
6.357488, 281
--------------------------------------人脸关键点检测任务 - tm_landmark.cpp
使用图片：
export LD_LIBRARY_PATH=./build/install/lib
./build/install/bin/tm_landmark -m models/landmark.tmfile -i images/mobileface02.jpg -r 1 -t 1结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
Repeat [1] min 17.461 ms, max 17.461 ms, avg 17.461 ms
ssd 目标检测任务 - tm_mobilenet_ssd.cpp
使用图片：
$ export LD_LIBRARY_PATH=./build/install/lib
$ ./build/install/bin/tm_mobilenet_ssd -m models/mobilenet_ssd.tmfile -i images/ssd_dog.jpg -r 1 -t 1结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
Repeat 1 times, thread 1, avg time 206.30 ms, max_time 206.30 ms, min_time 206.30 ms
--------------------------------------
detect result num: 3
dog :99.8%
BOX:( 138 , 209 ),( 324 , 541 )
car :99.7%
BOX:( 467 , 72 ),( 687 , 171 )
bicycle :99.5%
BOX:( 107 , 141 ),( 574 , 415 )
======================================
[DETECTED IMAGE SAVED]:
======================================
retinaface 人脸检测任务 - tm_refinaface.cpp
使用图片：
$ export LD_LIBRARY_PATH=./build/install/lib
$ ./build/install/bin/tm_retinaface -m models/retinaface.tmfile -i images/mtcnn_face4.jpg -r 1 -t 1结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
img_h, img_w : 316, 474
Repeat 1 times, thread 1, avg time 75.72 ms, max_time 75.72 ms, min_time 75.72 ms
--------------------------------------
detected face num: 4
BOX 0.99:( 38.9179 , 86.3346 ),( 45.7028 , 63.2934 )
BOX 0.99:( 168.12 , 86.14 ),( 37.5249 , 47.7839 )
BOX 0.98:( 383.673 , 56.4136 ),( 77.176 , 83.8093 )
BOX 0.98:( 289.365 , 103.773 ),( 38.0025 , 47.6989 )
yolact 实例分割任务 - tm_yolact.cpp
使用图片：
$ export LD_LIBRARY_PATH=./build/install/lib
$ ./build/install/bin/tm_yolact -m models/yolact.tmfile -i images/ssd_car.jpg -r 1 -t 1结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
Repeat 1 times, thread 1, avg time 15833.47 ms, max_time 15833.47 ms, min_time 15833.47 ms
--------------------------------------
6 = 0.99966 at 130.82 57.77 340.78 x 237.36
3 = 0.99675 at 323.39 194.97 175.57 x 132.96
1 = 0.33431 at 191.24 195.78 103.06 x 179.22
yolov3 目标检测任务 - tm_yolov3.cpp
使用图片：
$ export LD_LIBRARY_PATH=./build/install/lib
$ ./build/install/bin/tm_yolact -m models/yolov3_tiny.tmfile -i images/ssd_dog.jpg -r 1 -t 1结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
Repeat 1 times, thread 1, avg time 262.52 ms, max_time 262.52 ms, min_time 262.52 ms
--------------------------------------
num_detections,4
16: 57%
left = 129,right = 369,top = 186,bot = 516
2: 65%
left = 465,right = 677,top = 74,bot = 171
1: 60%
left = 205,right = 576,top = 153,bot = 447
yolov4-tiny目标检测任务 - tm_yolov4_tiny.cpp
使用图片：
$ export LD_LIBRARY_PATH=./build/install/lib
$ ./build/install/bin/tm_yolov4_tiny -m models/yolov4_tiny.tmfile -i images/ssd_dog.jpg -r 1 -t 1结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
Repeat 1 times, thread 1, avg time 177.72 ms, max_time 177.72 ms, min_time 177.72 ms
--------------------------------------
num_detections,10
16: 74%
left = 125,right = 327,top = 221,bot = 537
2: 40%
7: 84%
left = 455,right = 703,top = 77,bot = 168
1: 28%
left = 56,right = 603,top = 85,bot = 496
人体姿态识别任务 - tm_openpose.cpp
使用图片：
$ export LD_LIBRARY_PATH=./build/install/lib
$ ./build/install/bin/tm_openpose -m models/openpose_coco.tmfile -i image/pose.jpg -r 1 -t 1结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
Repeat 1 times, thread 1, avg time 15350.25 ms, max_time 15350.25 ms, min_time 15350.25 ms
--------------------------------------
KeyPoints Coordinate:
0:[292.174, 55.6522]
1:[306.087, 125.217]
2:[250.435, 139.13]
3:[236.522, 222.609]
4:[222.609, 306.087]
5:[361.739, 125.217]
6:[403.478, 208.696]
7:[417.391, 292.174]
8:[264.348, 306.087]
9:[264.348, 431.304]
10:[264.348, 570.435]
11:[347.826, 306.087]
12:[375.652, 431.304]
13:[333.913, 542.609]
14:[278.261, 41.7391]
15:[306.087, 41.7391]
16:[264.348, 55.6522]
17:[320, 55.6522]
人体姿态识别结果会保存为图片，名称为：Output-Keypionts.jpg和Output-Skeleton.jpg。
汉字识别任务 - tm_crnn.cpp
目录结构: 在根目录下新建model文件夹, 并从Tengine model zoo ( 密码：hhgc ) 中下载对应的模型文件和测试图片放入其中, 这里是: crnn_lite_dense.tmfile模型文件, o2_resize.jpg测试图片, 以及keys.txt字库文件. 在以上步骤的基础上, 输入如下命令行
测试图片：
export LD_LIBRARY_PATH=./build/install/lib
./build/install/bin/tm_crnn -m model/crnn_lite_dense.tmfile -i model/o2_resize.jpg -l model/keys.txt结果如下：
start to run register cpu allocator
tengine-lite library version: 1.0-dev
Repeat 1 times, thread 1, avg time 43.32 ms, max_time 43.32 ms, min_time 43.32 ms
--------------------------------------
如何突破自己的颜值上限
--------------------------------------其中ocr的识别结果会直接打印到终端中, 同时如果需要保存为txt文件可以修改源码使其重定向到文件。
我们将持续更新各种有趣的 demo ，敬请期待......