Towards Backward-Compatible Representation Learning
Abstract We propose a way to learn visual features that are compatible with previously computed ones even when they have different dimensions and are learned via different neural network architectures […]
Towards Backward-Compatible Representation Learning Read More »









