Closed MortironX closed 6 months ago
可以的,不过本模型是行为识别模型,要用预训练的话,需要用比如K400的动作标签
你好,我想完成一个手语识别,因此从博客中看到你们的代码,我有手语动作的视频和图片以及标签,想请问下,如果使用你们的模型你预估是否能达到预期的目标,完成手语识别。
好的,我进一步了解一下,感谢!
你好,我想请教一下,我根据我的手语识别任务,选择使用UMT的微调,目前需要修改data.py里数据路径,下面的cc3m模型,只需要下载umt-b 25m的一个模型就好了?最后想问一下该项目大约需要多大显存,感谢答复
获取Outlook for Androidhttps://aka.ms/AAb9ysg
From: Kunchang Li @.> Sent: Tuesday, April 9, 2024 8:44:36 PM To: OpenGVLab/UniFormerV2 @.> Cc: MortironX @.>; Author @.> Subject: Re: [OpenGVLab/UniFormerV2] 能否利用本模型完成其他的下游任务? (Issue #69)
可以基于本模型微调,或者基于我们更强的模型UMThttps://github.com/OpenGVLab/unmasked_teacher微调
― Reply to this email directly, view it on GitHubhttps://github.com/OpenGVLab/UniFormerV2/issues/69#issuecomment-2045098645, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BHWJHP7NJBIRNSFMTTIC5QLY4PPDJAVCNFSM6AAAAABF6GX522VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANBVGA4TQNRUGU. You are receiving this because you authored the thread.Message ID: @.***>
可以自己跑一下看看显存需求哈,直接用umt-b 25m预训练就好
好的,感谢
获取Outlook for Androidhttps://aka.ms/AAb9ysg
From: Kunchang Li @.> Sent: Friday, April 26, 2024 6:38:22 PM To: OpenGVLab/UniFormerV2 @.> Cc: MortironX @.>; Author @.> Subject: Re: [OpenGVLab/UniFormerV2] 能否利用本模型完成其他的下游任务? (Issue #69)
可以自己跑一下看看显存需求哈,直接用umt-b 25m预训练就好
― Reply to this email directly, view it on GitHubhttps://github.com/OpenGVLab/UniFormerV2/issues/69#issuecomment-2079132715, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BHWJHP62JIOZQUJIUTUAOX3Y7IVB5AVCNFSM6AAAAABF6GX522VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANZZGEZTENZRGU. You are receiving this because you authored the thread.Message ID: @.***>
你好,我想请问一下。现在我有一定的数据集(图片和视频)能否把我拥有的数据集喂给模型然后完成一定的识别任务,类似一个输入视频,输出文本的任务?