Visual Commonsense Representation Learning via Causal Inference

Tan Wang, Jianqiang Huang, Hanwang Zhang, Qianru Sun; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 378-379

We present a novel unsupervised feature representation learning method, Visual Commonsense Region-based Con-volutional Neural Network (VC R-CNN), to serve as an improved visual region encoder for high-level tasks such as captioning and VQA. Given a set of detected object regions in an image (e.g., using Faster R-CNN), like any other un-supervised feature learning methods (e.g., word2vec), the proxy training objective of VC R-CNN is to predict the con-textual objects of a region. However, they are fundamentally different: the prediction of VC R-CNN is by using causal intervention: P(Y|do(X)), while others are by using the conventional likelihood:P(Y|X). We extensively apply VC R-CNN features in prevailing models of two popular tasks: Image Captioning and VQA, and observe consistent performance boosts across all the methods, achieving many new state-of-the-arts. Code and feature are available at https://github.com/Wangt-CN/VC-R-CNN. For better clarity, you can also refer to the full version of this paper in [11].