diff --git a/TensorFlow/contrib/cv/resnet50/MyFirstApp/README.md b/TensorFlow/contrib/cv/resnet50/MyFirstApp/README.md new file mode 100644 index 0000000000000000000000000000000000000000..cbddbc392146f24626e635b1020db9ce12615a40 --- /dev/null +++ b/TensorFlow/contrib/cv/resnet50/MyFirstApp/README.md @@ -0,0 +1,72 @@ +## 模型功能 + + 对图像中的物体进行分类。 + +## 原始模型 + +参考实现 : + +https://github.com/KaimingHe/deep-residual-networks/tree/master/prototxt + +原始模型权重下载地址 : + +https://modelzoo-train-atc.obs.cn-north-4.myhuaweicloud.com/003_Atc_Models/AE/ATC%20Model/resnet50/resnet50.caffemodel + +原始模型网络下载地址 : + +https://modelzoo-train-atc.obs.cn-north-4.myhuaweicloud.com/003_Atc_Models/AE/ATC%20Model/resnet50/resnet50.prototxt + + +## om模型 + +om模型下载地址: + +https://modelzoo-train-atc.obs.cn-north-4.myhuaweicloud.com/003_Atc_Models/AE/ATC%20Model/resnet50/resnet50.om + +使用ATC模型转换工具进行模型转换时可以参考如下指令,具体操作详情和参数设置可以参考 [ATC工具使用指导](https://support.huaweicloud.com/ti-atc-A200dk_3000/altasatc_16_002.html) + +``` +atc --model=model/resnet50.prototxt --weight=model/resnet50.caffemodel --framework=0 --output=model/resnet50 --soc_version=Ascend310 +``` + +## 使用msame工具推理 + +参考 https://gitee.com/ascend/tools/tree/master/msame, 获取msame推理工具及使用方法。 + +获取到msame可执行文件之后,将待检测om文件放在model文件夹,然后进行性能测试。 + +## 性能测试 + +使用msame推理工具,发起推理性能测试。可以参考如下指令: + +``` + ./msame --model ../../../models/resnet50.om --output output/ --loop 100 + +``` + +性能测试数据为: + +``` +[INFO] output data success +Inference average time: 2.314100 ms +Inference average time without first time: 2.312818 ms +[INFO] destroy model input success +[INFO] unload model success, model Id is 1 +[INFO] Execute sample success +Test Finish! +``` + +平均推理性能 2.31ms + +## 精度测试 + +待完善 + +推理效果 +``` +top 1: index[161] value[0.764648] +top 2: index[162] value[0.156616] +top 3: index[167] value[0.038971] +top 4: index[163] value[0.021698] +top 5: index[166] value[0.011887] +``` \ No newline at end of file