评论 LeonL：发了楼主的后处理给你

[quote][size=2][url=forum.php?mod=redirect&goto=findpost&pid=479912&ptid=90381][color=#999999]回复：cjl68 发表于 2020-11-25 17:29[/color][/url][/size] 你好，Presenter包括Presenter Server与Presenter Agent。Presenter Agent提供一系列API，用户通过调用API[/quote]谢谢！我试试

这个链接图是花的啊 看不清楚

兄弟你解决了吗

yolov3的后处理这块代码逻辑，发我一份可以吗 谢谢 ! wyfida@163.com

[libprotobuf ERROR google/protobuf/wire_format_lite.cc:603] String field 'ge.proto.GraphDef.name' contains invalid UTF-8 data when parsing a protocol buffer. Use the 'bytes' type if you intend to send raw bytes.

您好， ERROR] GE(3438,atc):2020-10-27-23:23:57.120.724 [common/graph/./model.cc:167]3438 LoadFromFile: ErrorNo: -1(failed) ParseFromFileDescriptor failed [ERROR] GE(3438,atc):2020-10-27-23:23:57.120.789 [framework/domi/offline/main.cc:845]3438 GenerateModel: ErrorNo: -1(failed) Load model from ./x.pb failed, please check model file or input parameter[--framework] is correct 将模型转为onnx格式后将pb文件转为om时出现上述问题 请问是什么原因？环境下caffe的模型可以成功装换

这…… GPU跑出来的模型，到底能不能转为om放在晟腾310上面跑。

已经解决了

wang@magicbook:~/Downloads/2402/sample-lightsdetection$ tree -L 2.├── deploy.sh├── license.md├── lightsdetectionapp│   ├── camera_datasets│   ├── common│   ├── graph_deploy.config│   ├── graph_template.config│   ├── lightsdetectionapp.mind│   ├── lights_detection_inference│   ├── lights_detection_post_process│   ├── main.cpp│   ├── main.h│   ├── Makefile│   ├── out│   └── tool_api.h├── MyDataset│   └── camera_datasets├── MyModel│   ├── aipp_lights_detection.cfg│   ├── lights_detection│   ├── README.md│   ├── reassign_operators│   └── shape_lights_detection├── presenterserver│   ├── common│   ├── __init__.py│   ├── lights_detection│   ├── presenter_server.py│   ├── README│   └── requirements├── proto│   ├── ai_types.proto│   └── graph_config.proto├── README_cn.md├── README.md├── run_lightsdetectionapp.sh├── script│   ├── build_ezdvpp.sh│   ├── build_presenteragent.sh│   ├── build.sh│   ├── deploy_sdk.sh│   ├── ezdvpp│   ├── func_deploy.sh│   ├── func_util.sh│   ├── lights_detection.om│   ├── prepare_graph.sh│   ├── prepare_model.sh│   ├── prepare_param.sh│   ├── prepare_presenter_server.sh│   └── presenteragent├── stop_lightsdetectionapp.sh└── Third Party Open Source Software Notice .md日志在附件中 上面是整个人脸识别工程的目录结构 （我们只是修改了名字）我们的想法是只把detectionapp的内容整理出来 不使用presenter  现在presenter不启动，编译的可执行文件就无法正常执行