Closed SikaStar closed 4 years ago
Why the results of MMT you provided are much higher than the original paper?
In the original paper, MMT adopts the avg pooling in the model and k-means clustering method. We provide the results of MMT, which used the GeM pooling and DBSCAN clustering method.
Why the results of MMT you provided are much higher than the original paper?