Cross-Modal Scene Networks
This work addresses the challenge of scene recognition across different modalities, which is incremental as it builds on existing convolutional neural networks with regularization techniques.
The paper tackles the problem of learning cross-modal scene representations that transfer across modalities by introducing a new dataset and methods to regularize convolutional neural networks for a shared, modality-agnostic representation. The experiments show that this representation aids in cross-modal retrieval and visualizations reveal units activating on consistent concepts independently of modality.
People can recognize scenes across many different modalities beyond natural images. In this paper, we investigate how to learn cross-modal scene representations that transfer across modalities. To study this problem, we introduce a new cross-modal scene dataset. While convolutional neural networks can categorize scenes well, they also learn an intermediate representation not aligned across modalities, which is undesirable for cross-modal transfer applications. We present methods to regularize cross-modal convolutional neural networks so that they have a shared representation that is agnostic of the modality. Our experiments suggest that our scene representation can help transfer representations across modalities for retrieval. Moreover, our visualizations suggest that units emerge in the shared representation that tend to activate on consistent concepts independently of the modality.