Does the pretrained visual backbone exclude val/test images on refcoco/+/g datasets ?

[seanzhuh]

wget -P ext https://pjreddie.com/media/files/yolov3.weights

It seems that it actually does not exclude val/test images on refcoco/+/g datasets, this is unfair, since not removing val/test images during pretraining usually brings about 2-3 improments

[hbb1]

Thank you for your reminder. Yes, we haven't tried other pretrained backbone and this is somehow unfair compared to some two-stage methods. We may update the results using pretrained backbone exclude val/test images of refcoco series dataset. Actually, our setting follows one-stage methods like FAOA and ReSC with the same backbone.

[seanzhuh]

Thank you very much, still a great work. I'm wondering if you could kindly provide the ReferItGame dataset to me since the provided bash script can not download it, the link REFERIT_SPLITS_URL="https://s3-sa-east-1.amazonaws.com/query-objseg/referit_splits.tar.bz2" and REFERIT_DATA_URL="http://www.eecs.berkeley.edu/~ronghang/projects/cvpr16_text_obj_retrieval/referitdata.tar.gz" is down. Could you send this dataset to my e-mail, which is sean.zhuh@gmail.com.

[hbb1]

I've uploaded the missing data here. https://pan.baidu.com/s/18KkE3dpkffEoK3QU85uPnA
Verification code: fi9c

[hbb1]

Solved. So I am closing this issue. Please open more if other issues exist.