Commonly Uncommon: Semantic Sparsity in Situation Recognition

TL;DR

This paper addresses semantic sparsity in situation recognition by introducing a tensor composition method and web data augmentation, significantly improving structured prediction accuracy in complex visual tasks.

Contribution

It proposes a novel tensor composition function and semantic data augmentation techniques to better handle rare object-role combinations in situation recognition.

Findings

Tensor composition improves role-noun sharing and accuracy.

Web data augmentation enhances performance further.

Achieves up to 9.57% relative improvement in accuracy.

Abstract

Semantic sparsity is a common challenge in structured visual classification problems; when the output space is complex, the vast majority of the possible predictions are rarely, if ever, seen in the training set. This paper studies semantic sparsity in situation recognition, the task of producing structured summaries of what is happening in images, including activities, objects and the roles objects play within the activity. For this problem, we find empirically that most object-role combinations are rare, and current state-of-the-art models significantly underperform in this sparse data regime. We avoid many such errors by (1) introducing a novel tensor composition function that learns to share examples across role-noun combinations and (2) semantically augmenting our training data with automatically gathered examples of rarely observed outputs using web data. When integrated within a complete CRF-based structured prediction model, the tensor-based approach outperforms existing state of the art by a relative improvement of 2.11% and 4.40% on top-5 verb and noun-role accuracy, respectively. Adding 5 million images with our semantic augmentation techniques gives further relative improvements of 6.23% and 9.57% on top-5 verb and noun-role accuracy.