WXinlong / SOLO

SOLO and SOLOv2 for instance segmentation, ECCV 2020 & NeurIPS 2020.
Other
1.69k stars 307 forks source link

How to get the "acc", "AP" of each category , and the "mAP" #198

Open lijoe123 opened 2 years ago

lijoe123 commented 2 years ago

Could you please tell me how to get the "acc", "AP"of each category, and the mAP. Thank you !

DanielWuuuuuu commented 2 years ago

I have the same question!!! @WXinlong

DanielWuuuuuu commented 2 years ago

I got it! mmdet/core/evaluation/coco_utils.py, coco_eval, classwise👉True@lijoe123

tpfLZ commented 2 years ago

你好,能加下你的好友吗,我也想问下怎么具体操作得到solov2的精度@JakobWu

lijoe123 commented 2 years ago

mmdet/core/evaluation/coco_utils.py, coco_eval, classwise👉别人发给我的答案,您看看行不行。 发自坚果 R2

你好,能加下你的好友吗,我也想问下怎么具体操作得到solov2的精度

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications on the go with GitHub Mobile for iOS or Android.

[

{

@.***": "http://schema.org",

@.***": "EmailMessage",

"potentialAction": {

@.***": "ViewAction",

"target": "https://github.com/WXinlong/SOLO/issues/198#issuecomment-962390008",

"url": "https://github.com/WXinlong/SOLO/issues/198#issuecomment-962390008",

"name": "View Issue"

},

"description": "View this Issue on GitHub",

"publisher": {

@.***": "Organization",

"name": "GitHub",

"url": "https://github.com"

}

}

tpfLZ commented 2 years ago

您好,是直接运行这个py文件就行了吗,还是需要什么别的改动

lijoe123 commented 2 years ago

是把里面的classwise改为True发自坚果 R2

您好,是直接运行这个py文件就行了吗,还是需要什么别的改动

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications on the go with GitHub Mobile for iOS or Android.

[

{

@.***": "http://schema.org",

@.***": "EmailMessage",

"potentialAction": {

@.***": "ViewAction",

"target": "https://github.com/WXinlong/SOLO/issues/198#issuecomment-962393630",

"url": "https://github.com/WXinlong/SOLO/issues/198#issuecomment-962393630",

"name": "View Issue"

},

"description": "View this Issue on GitHub",

"publisher": {

@.***": "Organization",

"name": "GitHub",

"url": "https://github.com"

}

}

DanielWuuuuuu commented 2 years ago

解决了吗? @tpfLZ

tpfLZ commented 2 years ago

我还是没弄好,你能加我一下QQ:287558659吗?想问下具体怎么操作,非常感谢!@JakobWu

lijoe123 commented 2 years ago

你好,我也有问题了,我把classwise改为True 了,也不行,也是没有输出各个类别的AP值 @JakobWu

DanielWuuuuuu commented 2 years ago

@lijoe123 你好!你之前不是改好了吗?

lijoe123 commented 2 years ago

@JakobWu 我之前用了mmdetection,不是用这个,程序跑完还能输出AP值吗

lijoe123 commented 2 years ago

你好,我是直接在训练过程中就不显示评估结果 solov2

DanielWuuuuuu commented 2 years ago

@lijoe123 我记得好像是分布式训练中可以在训练中显示每一个epoch评估结果,非分布式训练中没有,但是我看之前的问题中好像分布式训练中显示评估结果也有点问题

lijoe123 commented 2 years ago

@lijoe123 我记得好像是分布式训练中可以在训练中显示每一个epoch评估结果,非分布式训练中没有,但是我看之前的问题中好像分布式训练中显示评估结果也有点问题

分布式训练是怎么实现的呢,可以给我个demo吗

DanielWuuuuuu commented 2 years ago

@lijoe123 我没有实现过,我只是知道在哪开启这个分布式训练,tools/train.py中第40-44行,75-80行,115-121行,你可以看看

blurmemo commented 1 year ago

@lijoe123 @DanielWuuuuuu 你好,请问这个你现在已经解决了吗?我想请教一下,谢谢

DanielWuuuuuu commented 1 year ago

@blurmemo 你指的是什么问题?

blurmemo commented 1 year ago

@blurmemo 你指的是什么问题?

抱歉我前段时间在复习考试,就是拿到模型的acc、AP、mAP和test Time这些指标

peiqilady commented 1 year ago

@blurmemo 你指的是什么问题?

抱歉我前段时间在复习考试,就是拿到模型的acc、AP、mAP和test Time这些指标

请问这个问题解决了吗

peiqilady commented 1 year ago

I got it! mmdet/core/evaluation/coco_utils.py, coco_eval, classwise👉True@lijoe123

我改成了true但是并没有变化