Is the preprocess method of GOT-10K and VID the same as SiamFC paper? about siamdw HOT 7 CLOSED

Thanks for your code fisrt, it help so much
Here is a question:
I used the same processing method as SiamFC paper mentioned to preprocess the GOT-10K, but when I tried to train SiamFCRes22 on it, the auc result of OTB2015 is only 0.61+

1） have you tuning hyper-parameter?
2) the default settings in SiamFC.yaml can give a pretty good results, especially when GOT10K is used.

from siamdw.

@herbert-chen did you use the author pre-process VID and got-10k??

from siamdw.

No, that google drive file is too large
@wangaixue

from siamdw.

can you share me your link? the link the author provided has the root @herbert-chen

from siamdw.

can you share me your link? the link the author provided has the root @herbert-chen

you can apply to access, and will be permitted sooner.

from siamdw.

No, that google drive file is too large
@wangaixue

the pairs are same as original SiamFC.

from siamdw.

Thanks for your code fisrt, it help so much
Here is a question:
I used the same processing method as SiamFC paper mentioned to preprocess the GOT-10K, but when I tried to train SiamFCRes22 on it, the auc result of OTB2015 is only 0.61+

We provide more training data now, including VID, YTB, COCO, DET, GOT10K and LASOT. And a pre-processing script is available. See train.md for details. This issue will be closed.

from siamdw.