Compositional Mixture Representations for Vision and Text
This addresses the challenge of interpretable cross-modal learning for computer vision and NLP researchers, but it is incremental as it builds on existing representation learning and spatial transformer methods.
The paper tackles the problem of learning a shared representation between vision and language without explicit location supervision, achieving weakly supervised object detection and extrapolation to unseen object combinations on MNIST and CIFAR10 variations.
Learning a common representation space between vision and language allows deep networks to relate objects in the image to the corresponding semantic meaning. We present a model that learns a shared Gaussian mixture representation imposing the compositionality of the text onto the visual domain without having explicit location supervision. By combining the spatial transformer with a representation learning approach we learn to split images into separately encoded patches to associate visual and textual representations in an interpretable manner. On variations of MNIST and CIFAR10, our model is able to perform weakly supervised object detection and demonstrates its ability to extrapolate to unseen combination of objects.