当前位置:网站首页>What's wrong with running yolov5 (1) p, r, map are all 0
What's wrong with running yolov5 (1) p, r, map are all 0
2022-08-02 14:19:00 【weixin_50862344】
1. I encountered an embarrassing problem when running today
During training, p, r, and map are all 0
The above image may not be very intuitive
Including the final result also has no precision, recall and map.
There are many reasons for no results after inquiring (the following reasons range from easy to operate to difficult to operate records):
①The format is not converted
②There is a problem with the folder format (name and structure)
A certainSome versions seem to have metaphysical renaming operations
③The pytorch and cuda versions do not correspond
This time I made a small mistake that only a wise man would make.A point to remind: When checking, you must start from the bottom.
My root cause is: txt for train training set, xml for val
Then I think there is a point that few bloggers will mention:
yolov5 actually only uses the corresponding image files and txt folders for training.In fact, as long as you have these two, it will be fine!!!
边栏推荐
- 监管再次重拳出击,后市如何?2021-05-22
- [ROS](03)CMakeLists.txt详解
- You can't accept 60% slump, there is no eligible for gain of 6000% in 2021-05-27
- [ROS] The difference between roscd and cd
- Some impressions of the 519 plummet 2021-05-21
- Object detection scene SSD-Mobilenetv1-FPN
- Sentinel源码(五)FlowSlot以及限流控制器源码分析
- 网络剪枝(1)
- AWVS工具介绍[通俗易懂]
- ftp常用命令详解_iftop命令详解
猜你喜欢
随机推荐
动手学ocr(一)
The world's largest Apache open source foundation is how it works?
[ROS] Compiling packages packages encounters slow progress or stuck, use swap
paddleocr window10 first experience
[ROS](01)创建ROS工作空间
Sentinel源码(三)slot解析
VMM是什么?_兮是什么意思
跑跑yolov5吧
网页设计(新手入门)[通俗易懂]
【Tensorflow】AttributeError: '_TfDeviceCaptureOp' object has no attribute '_set_device_from_string'
深度学习框架pytorch快速开发与实战chapter4
Configure zabbix auto-discovery and auto-registration.
第十四单元 视图集及路由
Flask上下文,蓝图和Flask-RESTful
MarkDown语法汇总
【ROS】工控机的软件包不编译
Flask框架
Flask请求应用上下文源码分析
第二届中国Rust开发者大会(RustChinaConf 2021~2022)线上大会正式开启报名
第五单元 保持状态