Summary
Graphic designs are made up of layers of different parts, like shapes and text, but most computer programs look at these parts as if they were all mixed up. This paper shows a new way to find and understand these parts by looking at them in the order they overlap, so that the bottom parts help to find the top parts better. They also predict the full size of each part, even the parts hidden behind others. To train this method, the authors created a huge collection of graphic designs and tested their approach, which worked better than previous methods and almost as well as humans at finding hidden parts. They also improved how the system learns by giving it more detailed feedback about each detected element.
graphic designobject detectioncompositional orderamodal detectionbounding boxautoregressive decodingpolicy optimizationmachine learninglayered elementsimage decomposition
Authors
Jiangning Zhu, Bowen Li, Shenyu Qiao, Yima Gu, Zhao Zhang, Yuhui Yuan, Shixia Liu
Abstract
Graphic designs, such as posters, advertisements, and infographics, are an important medium for communicating information and shaping understanding. Unlike natural images, they consist of layered elements with explicit compositional order. However, existing object detection models treat these elements as an unordered set, leaving compositional order unexploited. To address this limitation, we present Detect Anything in Graphic Design (DAD), a model that formulates graphic design detection as compositional deconstruction. It decodes elements in compositional order, using lower-layer elements to better detect higher-layer ones. The key feature of DAD is amodal detection, which predicts the full bounding box of each element, including regions occluded by elements placed above it. Building on this formulation, we propose Element Relative Policy Optimization (EleRPO), which extends GRPO from sequence-level supervision to element-level optimization. EleRPO provides fine-grained training signals that capture how each detected element contributes to overall detection quality, and works synergistically with compositional order to improve detection performance. To support training and evaluation, we build a dataset of 10 million graphic designs. Experiments show that DAD outperforms all baselines and achieves human-level performance in amodal detection, supporting effective image-to-layer decomposition. EleRPO consistently improves over GRPO across nine detection benchmarks.