运行如下命令，编译本项目： mkdir build && cd build && make all pdl 注：目前版本配置文件有硬编码，需按提示进行修改。
NCNN: onnx转ncnn命令： bin/ncnn/onnx2ncnn model_param/resnet18-v2-7.onnx model_param/ncnn/resnet18.param model_param/ncnn/resnet18.bin
运行推理命令： bin/ncnn/resnet18_ncnn image/dog.jpg
性能报告： === Performance Report (10 runs) === Min/Max/Avg Preprocessing Time: 0.10/0.11/0.10 ms Min/Max/Avg Inference Time: 27.23/57.88/33.95 ms Min/Max/Avg Postprocessing Time: 0.00/0.01/0.00 ms Memory Usage: Base=224612 KB, Peak=235828 KB, Growth=11216 KB Min/Max/Avg FPS: 17.24/36.58/29.36
Top 3 Results: Class: 188, Score: 9.829282 Class: 218, Score: 8.137074 Class: 157, Score: 7.802757
MNN: onnx转mnn命令： ./mnn/MNNConvert -f ONNX --modelFile ../model_param/resnet18-v2-7.onnx --MNNModel ../model_param/mnn/resnet18.mnn --bizCode biz
运行推理命令： bin/mnn/resnet18_mnn model_param/mnn/resnet18.mnn image/dog.jpg
性能报告： The device supports: i8sdot:0, fp16:0, i8mm: 0, sve2: 0 Input shape: [1, 3, 224, 224] Initial memory: 264968 KB
=== Performance Report (10 runs) === Min/Max/Avg Preprocessing Time: 0.65/2.70/0.91 ms Min/Max/Avg Inference Time: 19.64/38.58/22.75 ms Min/Max/Avg Postprocessing Time: 0.01/0.01/0.01 ms Memory Usage: Base=264968 KB, Peak=276900 KB, Growth=11932 KB Min/Max/Avg FPS: 24.22/49.16/44.05
Top 3 Results: Class: 188, Score: 9.5002 Class: 218, Score: 8.6247 Class: 215, Score: 8.5897
PaddleLite: onnx转paddle内部格式命令： x2paddle --framework=onnx --model=resnet18-v2-7.onnx --save_dir=pd_model 注：需先按照转换工具，转换后的文件保存至src/pdl/src/assets/models下
运行推理命令： cd src/pdl/src/shell && ./run.sh resnet imagenet_224.txt test linux amd64 cpu
性能报告： === Performance Report (10 runs) === Min/Max/Avg Preprocessing Time: 0.73/5.46/1.41 ms Min/Max/Avg Inference Time: 41.04/52.56/47.11 ms Min/Max/Avg Postprocessing Time: 0.85/2.69/1.19 ms Memory Usage: Base=178540 KB, Peak=195028 KB, Growth=16488 KB Min/Max/Avg FPS: 16.47/23.46/20.12
Top 3 Results: Class: 188, Score: 9.833639 Class: 218, Score: 8.139140 Class: 157, Score: 7.806087