search

MCG-NJU / MultiSports

[ICCV 2021] MultiSports: A Multi-Person Video Dataset of Spatio-Temporally Localized Sports Actions
Other
109 stars 7 forks source link

evaluate is failed #8

Open pokameng opened 3 years ago

pokameng commented 3 years ago

Hello @yixuanli98 I generate video_detections.pkl. it runs successfuly in my local machine,but it runs fail in online: only length-1 arrays can be converted to Python scalars I don not konw why

yixuanli98 commented 3 years ago

Maybe there are some errors in your data format. Also, save the pkl file with protocal=2

pokameng commented 3 years ago

I have set it =2。But  i  run my pkl successfully on my machine，fail online。But the plk is same！！！

---Original--- From: "Yixuan @.> Date: Sun, Sep 5, 2021 16:10 PM To: @.>; Cc: @.**@.>; Subject: Re: [MCG-NJU/MultiSports] evaluate is failed (#8)

Maybe there are some errors in your data format. Also, save the pkl file with protocal=2

— 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.

yixuanli98 commented 3 years ago

Actually the default python version of codalab is python2. You can test it on your machine with python2. Also, I give a submission example in this GitHub repository. You can check the format difference with your pkl file.

pokameng commented 3 years ago

I have compare it in type，but they are same。

------------------ 原始邮件 ------------------ 发件人: "MCG-NJU/MultiSports" @.>; 发送时间: 2021年9月5日(星期天) 下午4:19 @.>; @.**@.>; 主题: Re: [MCG-NJU/MultiSports] evaluate is failed (#8)

Actually the default python version of codalab is python2. You can test it on your machine with python2. Also, I give a submission example in this GitHub repository. You can check the format difference with your pkl file.

— 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.

yixuanli98 commented 3 years ago

It's strange. Other participants have submitted successfully and got satisfied results. How big is your submitted file?

yixuanli98 commented 3 years ago

The results of MOC are actually very large and my frame_detections.pkl is 8.2G for test set. If your baseline is MOC, the too large submit file may result in submission failure. Maybe you can have a try on the slowfast or design some strategies to reduce the predictions of MOC.

pokameng commented 3 years ago

My  pkls are 60M.i have sloved this problem,thanks！！！

---Original--- From: "Yixuan @.> Date: Mon, Sep 6, 2021 21:13 PM To: @.>; Cc: @.**@.>; Subject: Re: [MCG-NJU/MultiSports] evaluate is failed (#8)

The results of MOC are actually very large and my frame_detections.pkl is 8.2G for test set. If your baseline is MOC, the too large submit file may result in submission failure. Maybe you can have a try on the slowfast or design some strategies to reduce the predictions of MOC.

— 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.