prepare data problem about sc-sfmlearner-release HOT 3 CLOSED

I use python3.6. Indeed, this code is copied from Clement Pinard's pytorch implementation of SfMLearner. I tried it many times, and it works well. Maybe the order is not critical. For more details on data construction, I suggest you opening issues on his repo.

from sc-sfmlearner-release.

Thanks for your reply. I am very interesting in this work. I want to verify that the 'train.txt' and 'val.txt' differ when you prepare data every time. So it may result in the training set is different from other methods, like Sfmlearner and the np.random.seed(8964). may be useless. And we do not use 'val.txt' in the training process. The result works well every time maybe because the 'train.txt' is similar every time. I will continue to follow your this setting.

from sc-sfmlearner-release.

Thanks for your attention. Look forward to your new conclusions about the splits.

from sc-sfmlearner-release.