Open MarStarck opened 5 months ago
Yes, check the home page -- "Note that it may not be safe to train / test across different versions of the dataset, for example the k400 validation set is largely part of the k700 training set so results will not be meaningful."
Yes, check the home page -- "Note that it may not be safe to train / test across different versions of the dataset, for example the k400 validation set is largely part of the k700 training set so results will not be meaningful."