Recognizing Families In the Wild (FIW) Data Challenge Workshop - ACM MM 2017 Forum

Go back to competition Back to thread list Post in this thread

> Training Data Update

As several of you have brought up, there are some images in the training directory that are not in the training list. Reasons for this vary-- in some cases, there are distance relationships; in some others, we may have missed excluded them accidentally.

In any case, ALL IMAGES IN TRAINDATA can be used for training. Additionally, validation data can also be used to train the final models used for "blind" testing come towards the end (but that is slightly a different point).

To avoid confusion another file was added that contains all images in traindata (https://1drv.ms/u/s!AkDk_XdBkJ9wgwt15ckl3cPp9WqR), which will also be added to "Get Data" tab of website.

The new list is named train_full.list not to get anyone mixed up with train.list (which still is available as well).

Difference: train_full.list has 7,220 face images listed, while train.list has 6,111.

Let me know if you have any questions, suggestions, and or requestions.

Best regards,
Joe

Posted by: jvision @ June 15, 2017, 2:35 a.m.
Post in this thread