FED-Bench: A Cross-Granular Benchmark for Disentangled Evaluation of Facial Expression Editing

TL;DR

FED-Bench introduces a comprehensive, multi-dimensional benchmark and evaluation protocol for facial expression editing, addressing existing gaps in quality, instruction adherence, and bias mitigation.

Contribution

It provides a new scalable benchmark with a detailed evaluation suite and demonstrates its utility by improving model performance through additional training data.

Findings

Current models struggle with high-fidelity, accurate expression editing.

FED-Score effectively disentangles evaluation dimensions, reducing bias.

Fine-grained instruction following is the main bottleneck in current approaches.

Abstract

Facial expression image editing requires fine-grained control to strictly preserve human identity and background while precisely manipulating expression. However, existing editing benchmarks primarily focus on general scenarios, lacking high-quality facial images and corresponding editing instructions. Furthermore, current evaluation metrics exhibit systemic biases in this task, often favoring lazy editing or overfit editing. To bridge these gaps, we propose FED-Bench, a comprehensive benchmark featuring rigorous testing and an accurate evaluation suite. First, we carefully construct a benchmark of 747 triplets through a cascaded and scalable pipeline, each comprising an original image, an editing instruction, and a ground-truth image for precise evaluation. Second, we introduce FED-Score, a cross-granularity evaluation protocol that disentangles assessment into three dimensions: Alignment for verifying instruction following, Fidelity for testing image quality and identity preservation, and Relative Expression Gain for quantifying the magnitude of expression changes, effectively mitigating the aforementioned evaluation biases. Third, we benchmark 18 image editing models, revealing that current approaches struggle to simultaneously achieve high fidelity and accurate expression manipulation, with fine-grained instruction following identified as the primary bottleneck. Finally, leveraging the scalable characteristic of introduced benchmark engine, we provide a 20k+ in-the-wild facial training set and demonstrate its effectiveness by fine-tuning a baseline model that achieves significant performance gains. Our benchmark and related code will be made publicly open soon.