I am wondering if the bounding box will be furnished in the test set ?
Posted by: dahmanmo @ March 16, 2016, 2:08 p.m.Hi,
I'm sorry, could you clarify what do you mean by furnished?
Thank you.
Kind Regards,
Mercedes
Posted by: MTT @ March 27, 2016, 6:01 p.m.I want to say if we can use the bounding boxes as entry.
Posted by: dahmanmo @ March 27, 2016, 6:12 p.m.Hi,
Sorry, I don't think I understand. What do you mean entry? You mean use our bounding boxes in your training? Or generating your own bounding boxes?
Your prediction file should include the coordinates, width and height of the bounding boxes around the face that you have detected along with your classification.
Let me know if you meant something else.
Thank you.
Kind Regards,
Mercedes
Posted by: MTT @ March 27, 2016, 6:21 p.m.Is it true that the bounding boxes are not a task of the challenge neither for evaluation. So can I use the bounding boxes withing the training file.
Posted by: dahmanmo @ March 27, 2016, 6:26 p.m.Hi,
You may use the bounding boxes we provide in your training, of course. However, please bear in mind that these were obtained by our annotators, so they are not the ground-truth.
Your model should provide the bounding box around the testing data, but we will not evaluate that. We will evaluate just the prediction.
Thank you.
Kind Regards,
MTT
Posted by: MTT @ March 27, 2016, 6:41 p.m.Hi,
We will get bounding box information for the test set?
Thanks
Posted by: dpali @ March 28, 2016, 9:17 p.m.Hi,
The bounding boxes around the testing set will not be made public. That is why your model should generate them as part of your prediction file.
Thanks and good luck!
Kind Regards,
MTT
Posted by: MTT @ March 28, 2016, 9:20 p.m.It's okay if they are not public, but would they still be available like training set?
Can we use them as part of our pipeline assuming that will be available with you when you test our result?
Thanks
Posted by: dpali @ March 28, 2016, 9:25 p.m.It's okay if they are not public, but would they still be available like training set?
Can we use them as part of our pipeline assuming that will be available with you when you test our result?
Thanks
Posted by: dpali @ March 28, 2016, 9:25 p.m.What happens if there are multiple faces in the image and no bounding box information? Do we detect for all the faces in the image?
Thanks
Posted by: dpali @ March 28, 2016, 9:58 p.m.Hi,
In the event that multiple faces are detected within an image, choose the face closer to the centre. Each image should only have one prediction.
Thank you.
Kind Regards,
MTT
Posted by: MTT @ March 28, 2016, 10:01 p.m.