Auto-Encoding Scene Graphs for Image Captioning

TLDR

Humans use inductive bias to compose collocations and infer context, so incorporating such bias as a language prior can reduce overfitting in encoder‑decoder captioning models. The paper proposes a Scene Graph Auto‑Encoder that injects language inductive bias into encoder‑decoder image captioning to produce more human‑like captions. The approach models images and sentences as scene graphs, learns a shared dictionary that encodes language priors, and uses the S→G→D→S and I→G→D→S pipelines to transfer inductive bias across domains. On the MS‑COCO benchmark, the SGAE single model achieves a new state‑of‑the‑art 127.8 CIDEr‑D on the Karpathy split and 125.5 CIDEr‑D (c40) on the official server, rivaling ensemble models. Code is available at https://github.com/yangxuntu/SGAE.

Abstract

We propose Scene Graph Auto-Encoder (SGAE) that incorporates the language inductive bias into the encoder-decoder image captioning framework for more human-like captions. Intuitively, we humans use the inductive bias to compose collocations and contextual inference in discourse. For example, when we see the relation "person on bike'', it is natural to replace "on'' with "ride'' and infer "person riding bike on a road'' even the "road'' is not evident. Therefore, exploiting such bias as a language prior is expected to help the conventional encoder-decoder models less likely to overfit to the dataset bias and focus on reasoning. Specifically, we use the scene graph - a directed graph (G) where an object node is connected by adjective nodes and relationship nodes - to represent the complex structural layout of both image (I) and sentence (S). In the textual domain, we use SGAE to learn a dictionary (D) that helps to reconstruct sentences in the S → G → D → S pipeline, where D encodes the desired language prior; in the vision-language domain, we use the shared D to guide the encoder-decoder in the I → G → D → S pipeline. Thanks to the scene graph representation and shared dictionary, the inductive bias is transferred across domains in principle. We validate the effectiveness of SGAE on the challenging MS-COCO image captioning benchmark, e.g., our SGAE-based single-model achieves a new state-of-the-art 127.8 CIDEr-D on the Karpathy split, and a competitive 125.5 CIDEr-D (c40) on the official server even compared to other ensemble models. Code has been made available at: https://github.com/yangxuntu/SGAE.