Semantically Stable Image Composition Analysisvia Saliency and Gradient Vector Flow Fusion

2026-04-14

AI Summary

This work addresses the challenge of simultaneously achieving semantic robustness and high-quality visual layout assessment in image composition evaluation by proposing VFCNet. The method introduces a novel approach that fuses saliency and edge information into a Gradient Vector Flow (GVF) field, leveraging a dual-stream GVF representation, attention-based fusion, and multi-scale features extracted from a DINOv3 self-supervised backbone to construct composition representations robust to low-level semantic variations. Remarkably, even a simple classifier using only DINOv3 features outperforms existing sophisticated task-specific models, achieving state-of-the-art performance on the PICD benchmark with CDA-1 and CDA-2 scores of 0.683 and 0.629, respectively—improvements of 33.1% and 36.1% over the previous best method.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The reliable computational assessment of photographic composition requires features that are discriminative of spatial layout yet robust to semantic content. This paper proposes a low-level representation grounded in the assumption that composition can be understood as the flow of visual attention across geometric structure. We introduce VFCNet, which fuses saliency and edge information into a gradient vector flow (GVF) field. The model computes dual-stream GVF representations, integrates them via attention, and extracts multi-scale flow features with a DINOv3 backbone. VFCNet achieves state-of-the-art performance on the PICD benchmark (CDA-1: 0.683, CDA-2: 0.629), improving by 33.1\% and 36.1\% over the previous best method. We also show that a simple classifier on self-supervised DINOv3 features substantially outperforms more sophisticated, composition-specialized models. Code is available at https://github.com/ADadras/VFCNet

Problem

• image composition
• semantic robustness
• saliency
• gradient vector flow
• computational aesthetics

Innovation

• Gradient Vector Flow
• Saliency Fusion
• Visual Composition
• DINOv3
• Self-supervised Representation