上板运行提示json错误

根据大神“Ethanyyy”提供的上板教程把yolov5编译成功并部署到开发板上,运行"run_example.sh"脚本进行1080P的视频推理出现以下错误:

fb_cfg_func::platform is x3

fb_cfg_func::fb_res is 1080_fb

fb_cfg_func::vio_cfg_file is ./configs/x3_video_source.json.fb

sh: write error: No such device

/usr/bin/adbd is not runnning

Detecting platform:

board : Hobot XJ3 SOC SDB

udc : b2000000.dwc3

usb-gadget is already running

after start uvc

resolution 1080P feedback start, 1080P X3 JPEG Codec…

vio_cfg_file: ./configs/x3_video_source.json.fb

(video_source.cc:286): log level has been set, log_level: 4

(vps_manager.cc:375): create vps manager fifo success, this is main process

(vps_manager.cc:436): write input group id to fifo success, write_bytes_num: 32 fifo_fd: 5

(video_source_produce.cc:99): log level has been set, log_level: 4

(inferencer.cc:107): [ERROR] Not Support model_post_process-class_name: yolov5_postprocess

(method_manager.cc:112): methods init failed

错误日志已上传至附件

1642071604857.txt

(inferencer.cc:107): [ERROR] Not Support model_post_process-class_name: yolov5_postprocess

您好,后处理出错了,需要对应好输入输出接口

后处理这里已经处理过了,但是不知道为什么提示找不到class_name

代码是以库的形式提供的,需要对源码进行编译,生成库替换现有的库

代码是编译过的,生成的"libmodel_inference.so"也替换了到部署生成的deploy/lib目录了

需要内部加一些log 判断下,是否使用的是自己编译的库文件

问题已处理,手动拷贝model_inference.so/build目录下的libmodel_inference.so替换到生成的部署deploy/lib下,model_inference的make.sh 里面cp命令可能路径不对还是其他问题

现在新的问题是提示找不到模型了

[000:000] (packed_model.cpp:440): Model not exists: yolov5

(inference_engine_dnn.cc:203): not found model: yolov5

(inferencer.cc:128): LoadModelHandle failed.

(method_manager.cc:112): methods init failed

日志的最后输出:

[61971.364385] x3sdbx3-samsung2G-3200:unkown: Untracked pid 4420 killed by signal 9

[61971.364731] x3sdbx3-samsung2G-3200:unkown: Untracked pid 4421 killed by signal 9

[61979.829420] alloc_contig_range: [7f070, 7f080) PFNs busy

[61979.829729] module id: 4, event id: 1 register err[had registed]

[61979.834162] alloc_contig_range: [7f070, 7f080) PFNs busy

[61979.834286] module id: 4, event id: 2 register err[had registed]

[62533.646948] x3sdbx3-samsung2G-3200:unkown: Untracked pid 4495 killed by signal 9

[62533.647294] x3sdbx3-samsung2G-3200:unkown: Untracked pid 4496 killed by signal 9

[62542.287638] module id: 4, event id: 1 register err[had registed]

[62542.292030] alloc_contig_range: [7f080, 7f090) PFNs busy

[62542.292267] module id: 4, event id: 2 register err[had registed]

[62610.358053] x3sdbx3-samsung2G-3200:unkown: Untracked pid 4589 killed by signal 9

[62610.358389] x3sdbx3-samsung2G-3200:unkown: Untracked pid 4590 killed by signal 9

[62622.538675] alloc_contig_range: [7f070, 7f080) PFNs busy

[62622.538817] module id: 4, event id: 1 register err[had registed]

[62622.543416] alloc_contig_range: [7f070, 7f080) PFNs busy

[62622.543669] module id: 4, event id: 2 register err[had registed]

infer_yolov5_config.json配置如下:

{-
“mode”: “run_model”,-
“with_postprocess”: true,-
“model_preprocess”:{-
“class_name”: “pyramid_preprocess”,-
“pyramid_layer”: 8,-
“config”: {-
“image_process_pipeline”: [-
“pad(960, 960)”,-
“resize(672, 672)”-
]-
}-
},-
“model_predict”: {-
“model_file_path”: “/userdata/deploy/yolov5_mobilenetv2_example/models/yolov5_672x672_nv12.bin”,-
“is_packed_model”: false,-
“model_name”: “yolov5”,-
“run_mode”: {-
“bpu_core”:2-
}-
},-
“model_post_process”: {-
“class_name”: “yolov5_postprocess”,-
“score_threshold”: 0.3,-
“nms_threshold”: 0.45,-
“nms_top_k”: 500,-
“basic_pyramid_image_height”: 540,-
“basic_pyramid_image_width”: 960,-
“src_image_height”: 2160,-
“src_image_width”: 3840-
}-
}

反复测试过,模型的文件路径是没有问题的,提示

(inference_engine_dnn.cc:203): not found model: yolov5

(inferencer.cc:128): LoadModelHandle failed.

(method_manager.cc:112): methods init failed