About Test Data

39 views
Skip to first unread message

Vinh Lã Tuấn

unread,
May 23, 2021, 12:15:48 AM5/23/21
to MMSys'21 Grand Challenge on Detecting Cheapfakes
Hi, I want to ask that when evaluate proposal in challenge, can I use 'caption_modified' to detect if it's out of context or not(or should I use only image and caption only)?
I also want to ask which of attribute in test dataset I can use (maskrcnn_boxes, bert base/large score) when testing.

shivangi.tum

unread,
May 23, 2021, 5:57:06 AM5/23/21
to MMSys'21 Grand Challenge on Detecting Cheapfakes
Hi,

The hidden test set is exactly similar to the public test set in terms of attributes, so you are free to use any of the attributes, however, you need to provide us your exact solution so that we can replicate your results.
In short, yes you could use all of caption_modified, maskrcnn_boxes, bert base/large score at test, but make sure you include exact same steps while submitting.

Message has been deleted

Vinh Lã Tuấn

unread,
May 24, 2021, 8:54:54 PM5/24/21
to MMSys'21 Grand Challenge on Detecting Cheapfakes
So inshort in my source code while submitting, I should find Mask_RCNN bounding box, caption_modified myself right? Sorry for my dummy question but I just want to make everything clear.


Vào lúc 16:57:06 UTC+7 ngày Chủ Nhật, 23 tháng 5, 2021, shivangi.tum đã viết:

shivangi.tum

unread,
May 25, 2021, 5:46:25 AM5/25/21
to MMSys'21 Grand Challenge on Detecting Cheapfakes
yes you would find all the attributes that you see in the public test file. :)
Reply all
Reply to author
Forward
0 new messages