The person id of AVA

[jiagang zhu]

Thanks for sharing this dataset. 

I tried to visualize the person id of each person roi and I found many mismatching cases of the same person between different frames. 

I am wondering how the person ids of AVA were annotated. Thank you.

[David Ross]

Hi Jiagang Zhu,

Thanks for your interest in AVA. The method for annotating person ids is described in section 3.4 of the paper https://arxiv.org/pdf/1705.08421.pdf.

Are the problems that you're seeing false positives (two different people get the same id) or false negatives (the same person gets two different ids and is not linked)?

Regards,

David

--
You received this message because you are subscribed to the Google Groups "AVA Dataset Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ava-dataset-us...@googlegroups.com.
To post to this group, send email to ava-data...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ava-dataset-users/8eec3ac6-bd44-4be9-8833-8731c894aeba%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[jiagang zhu]

Thanks for your response.

I mean the latter false negatives. It seems that the shot cut of AVA movie and abrupt appearance change of person frequently change the person id of the same person.

在 2019年6月12日星期三 UTC+8上午2:11:51，David Ross写道：

Hi Jiagang Zhu,

Thanks for your interest in AVA. The method for annotating person ids is described in section 3.4 of the paper https://arxiv.org/pdf/1705.08421.pdf.

Are the problems that you're seeing false positives (two different people get the same id) or false negatives (the same person gets two different ids and is not linked)?

Regards,

David

On Sun, Jun 9, 2019 at 3:02 AM jiagang zhu <zhujia...@gmail.com> wrote:

Thanks for sharing this dataset. 

I tried to visualize the person id of each person roi and I found many mismatching cases of the same person between different frames. 

I am wondering how the person ids of AVA were annotated. Thank you.

--
You received this message because you are subscribed to the Google Groups "AVA Dataset Users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to ava-data...@googlegroups.com.

[David Ross]

Hi Jiagang Zhu,

I understand. Unfortunately these missed linkages are a limitation of the approach we used -- trying to connect people only across adjacent keyframes. Our focus is to identify atomic actions over short time stretches, and to possibly leverage short sequences of actions within a single shot to improve performance. Modeling extended action sequences across shot boundaries in AVA is an open problem.

Regards,

David

To unsubscribe from this group and stop receiving emails from it, send an email to ava-dataset-us...@googlegroups.com.

To post to this group, send email to ava-data...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/ava-dataset-users/e2c1fbe5-d452-487b-85d5-06073b024629%40googlegroups.com.