diff --git a/ACL_TensorFlow/contrib/cv/DeepSort_ID0505_for_ACL/README.md b/ACL_TensorFlow/contrib/cv/DeepSort_ID0505_for_ACL/README.md new file mode 100644 index 0000000000000000000000000000000000000000..50b9a505a73ef6a3bc6fd09d5ad20950b2afa52c --- /dev/null +++ b/ACL_TensorFlow/contrib/cv/DeepSort_ID0505_for_ACL/README.md @@ -0,0 +1,96 @@ +## 模型功能 + +目标跟踪 + +## 原始模型 + +参考: + + +原实现模型: + +https://gitee.com/ascend/ModelZoo-TensorFlow/tree/master/TensorFlow/contrib/cv/DeepSort_ID0505_for_TensorFlow + +pb文件下载地址 : + +链接:https://pan.baidu.com/s/1v7Fe_YYT-hZUGCf2u7TI5g +提取码:v41g + +## om模型 + +pb模型转om模型参考freeze_pb.py + + +om模型下载地址: + +链接:https://pan.baidu.com/s/1GCuGdEUiniYZlTdFn2sknw +提取码:keil + +使用ATC模型转换工具进行模型转换时可以参考如下指令: + +``` +atc --model=/root/deepsort/deep_sort.pb --framework=3 --output=/root/deepsort/deep_sort --soc_version=Ascend310 --input_shape="input:1,128,64,3" +``` + +## 数据集准备 + +market1501测试集中的图像数据转换为bin数据集 地址: + + +链接:https://pan.baidu.com/s/1h9bWDVEW7-voFHFi7TaTuw 提取码:nwyj + + +## 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,进行性能测试。 + +./msame --model "/root/deepsort/deep_sort.om" --input "/root/osnet/query" --output "/root/deepsort/out/" --outfmt TXT + + +## 性能测试 + +使用msame推理工具,参考如下命令,发起推理性能测试: + +``` + +``` + +``` +... +Inference average time : 10.66 ms +Inference average time without first time: 10.66 ms +[INFO] unload model success, model Id is 1 +[INFO] Execute sample success +[INFO] end to destroy stream +[INFO] end to destroy context +[INFO] end to reset device is 0 +[INFO] end to finalize acl + +... +``` + +平均推理性能为 10.66ms + +## 精度测试 + + +``` + +``` + +最终精度:(暂无) + +``` +Ascend310推理结果: + gpu结果: + npu结果: +``` + + + + +