【寒武纪硬件产品型号】必填*：MLU220edge

【使用操作系统】必填*：ubuntu20.04

【使用驱动版本】必填*：CNRT: 4.10.7 a16cc83”

【出错信息】必填*：我按文档配置cnstream的json文件如下：

未过滤配置：====================================================================================================

{

  "detector" : {

    "class_name" : "cnstream::Inferencer2",

    "parallelism" : 2,

    "max_input_queue_size" : 20,

    "custom_params" : {

      "model_path" : "../../../data/models/yolov5_b4c4_rgb_mlu220.cambricon",

      "func_name" : "subnet0",

      "preproc_name" : "VideoPreprocYolov5",

      "postproc_name" : "VideoPostprocYolov5",

      "keep_aspect_ratio" : "true",

      "model_input_pixel_format" : "RGB24",

      "batching_timeout" : 100,

      "threshold" : 0.6,

      "engine_num" : 2,

      "infer_interval" : 1,

      "device_id" : 0

    }

  }

}

标签未换，这个问题不大

过滤配置===========================================================================================================

{

  "detector" : {

    "class_name" : "cnstream::Inferencer2",

    "parallelism" : 2,

    "max_input_queue_size" : 20,

    "custom_params" : {

      "model_path" : "../../../data/models/yolov5_b4c4_rgb_mlu220.cambricon",

      "func_name" : "subnet0",

      "preproc_name" : "VideoPreprocYolov5",

      "postproc_name" : "VideoPostprocYolov5",

      "keep_aspect_ratio" : "true",

      "model_input_pixel_format" : "RGB24",

      "batching_timeout" : 100,

      "threshold" : 0.6,

      "engine_num" : 2,

      "infer_interval" : 1,

      " _infer" : "true",                                // 设置过滤

      "obj_filter_name" : "CarFilter",                    // 设置汽车过滤

      "device_id" : 0

    }

  }

}

控制台信息：

运行的命令：../../bin/cns_launcher --data_path ../../files.list_video --src_ _rate 25 --config_fname ./config.json --log_to_stderr=true

CNRT: 4.10.7 a16cc83

WARNING: Logging before InitCNStreamLogging() is written to STDERR

CNSTREAM REFLEX_  I0628 17:08:04.078861 177549] Register   named [PreprocLprnet]

CNSTREAM REFLEX_  I0628 17:08:04.079609 177549] Register   named [PreprocCpu]

CNSTREAM REFLEX_  I0628 17:08:04.079747 177549] Register   named [ObjPreprocCpu]

CNSTREAM REFLEX_  I0628 17:08:04.079860 177549] Register   named [PreprocYolov3]

CNSTREAM REFLEX_  I0628 17:08:04.079983 177549] Register   named [PreprocYolov5]

CNSTREAM REFLEX_  I0628 17:08:04.080096 177549] Register   named [VideoPreprocCpu]

CNSTREAM REFLEX_  I0628 17:08:04.080240 177549] Register   named [VideoObjPreprocCpu]

CNSTREAM REFLEX_  I0628 17:08:04.080358 177549] Register   named [VideoPreprocYolov3]

CNSTREAM REFLEX_  I0628 17:08:04.080482 177549] Register   named [VideoPreprocYolov5]

CNSTREAM REFLEX_  I0628 17:08:04.080632 177549] Register   named [PostprocClassification]

CNSTREAM REFLEX_  I0628 17:08:04.080775 177549] Register   named [ObjPostprocClassification]

CNSTREAM REFLEX_  I0628 17:08:04.080891 177549] Register   named [PostprocLprnet]

CNSTREAM REFLEX_  I0628 17:08:04.081015 177549] Register   named [PostprocMSSDPlateDetection]

CNSTREAM REFLEX_  I0628 17:08:04.081126 177549] Register   named [PostprocSsd]

CNSTREAM REFLEX_  I0628 17:08:04.081261 177549] Register   named [PostprocVehicleCts]

CNSTREAM REFLEX_  I0628 17:08:04.081399 177549] Register   named [PostprocYolov3]

CNSTREAM REFLEX_  I0628 17:08:04.081511 177549] Register   named [PostprocYolov5]

CNSTREAM REFLEX_  I0628 17:08:04.081660 177549] Register   named [VideoPostprocClassification]

CNSTREAM REFLEX_  I0628 17:08:04.081797 177549] Register   named [VideoObjPostprocClassification]

CNSTREAM REFLEX_  I0628 17:08:04.081915 177549] Register   named [VideoPostprocSsd]

CNSTREAM REFLEX_  I0628 17:08:04.082049 177549] Register   named [VideoPostprocYolov3]

CNSTREAM REFLEX_  I0628 17:08:04.082166 177549] Register   named [VideoPostprocYolov3MM]

CNSTREAM REFLEX_  I0628 17:08:04.082304 177549] Register   named [VideoPostprocYolov5]

CNSTREAM REFLEX_  I0628 17:08:04.082415 177549] Register   named [CarFilter]

CNSTREAM REFLEX_  I0628 17:08:04.082520 177549] Register   named [PlateFilter]

CNSTREAM REFLEX_  I0628 17:08:04.082634 177549] Register   named [VehicleFilter]

CNSTREAM REFLEX_  I0628 17:08:04.083153 177549] Register   named [PostprocBody25Pose]

CNSTREAM REFLEX_  I0628 17:08:04.083294 177549] Register   named [PostprocCOCOPose]

CNSTREAM REFLEX_  I0628 17:08:04.083635 177549] Register   named [DefaultKafkaHandler]

CNSTREAM DEMO I0628 17:08:04.083973 177549] CNSTREAM VERSION:v6.2.0

WARNING: Logging before InitGoogleLogging() is written to STDERR

I0628 17:08:04.198666 177549 mlu_context.cpp:130] [EasyDK Device] [CnrtInitTool] Cambricon runtime init success.

I0628 17:08:04.203487 177549 model_manager.cpp:172] [EasyDK InferServer] [ModelManager] Load model from file: ./../configs/../../../data/models/yolov5_b4c4_rgb_mlu220.cambricon

I0628 17:08:04.328605 177549 model_cnrt.cpp:213] [EasyDK InferServer] [Model] Load function from offline model succeeded

CNSTREAM CORE I0628 17:08:04.394804 177549] Pipeline[MyPipeline] Start

CNSTREAM CORE I0628 17:08:04.396076 177549] [stream_0]: Stream opening...

CNSTREAM CORE I0628 17:08:04.396694 177549] Add stream success, stream id : [stream_0]

CNSTREAM SOURCE I0628 17:08:04.450706 177594] [stream_0]: Got video info.

CNSTREAM SOURCE I0628 17:08:04.451182 177594] [stream_0]: Begin create decoder

CNSTREAM SOURCE I0628 17:08:04.473217 177594] [stream_0]: Finish create decoder

CNSTREAM VideoEncoderMlu I0628 17:08:04.724020 177592] VideoEncoderMlu200(1920x1080, NV12, H264)

CNSTREAM RtspServer I0628 17:08:04.726212 177600]  Stream URL "rtsp://192.168.1.60:9554/live"

CNSTREAM VideoEncoderMlu I0628 17:08:04.850968 177602] ReceivePacket() got parameter sets, size=36

**********************  Performance Print Start  (Whole)  **********************

===========================  Pipeline: [MyPipeline]  ===========================

---------------------- Module: [MyPipeline/decode/source] ----------------------

----------Process Name: [PROCESS]

[Counter]: 35, [Throughput]: 18.9955fps

---------------- Module: [MyPipeline/ _detection/detector] ----------------

----------Process Name: [PROCESS]

[Counter]: 35, [Throughput]: 359.321fps

------------------ Module: [MyPipeline/osd_label_map_coco/osd] -----------------

----------Process Name: [PROCESS]

[Counter]: 32, [Throughput]: 67.1891fps

--------------------- Module: [MyPipeline/sinker/rtsp_sink] -------------------- (slowest)

----------Process Name: [PROCESS]

[Counter]: 20, [Throughput]: 12.7911fps

-----------------------------------  Overall  ----------------------------------

[Counter]: 20, [Throughput]: 10.8604fps

***********************  Performance Print End  (Whole)  ***********************

CarFilter的代码：

【当前已做了哪些信息确认】选填：

确认是按文档一步步做的，而且都是官方代码和素材。

我的初衷是想要做区域过滤，只对感兴趣的区域做推理，所以试了下obj filter，发现不行。如果可行，是不是我把CarFilter中的代码换成区域过滤代码，返回true即可实现区域过滤？

希望大牛帮我解决一下。谢谢！

【参考配置文档链接】：基于推理服务的推理模块 — 寒武纪CNStream用户手册 6.2.0 文档 (cambricon.com)

【相关日志文档】选填
如有，可附件

【出错代码链接】选填：
github的或gitee的代码的链接，