Preserve or Modify? Context-Aware Evaluation for Balancing Preservation and Modification in Text-Guided Image Editing

TL;DR

AugCLIP is a novel context-aware evaluation metric for text-guided image editing that adaptively balances preservation and modification by modeling ideal edits in CLIP space, aligning closely with human judgment.

Contribution

This paper introduces AugCLIP, the first adaptive, context-aware metric that effectively balances preservation and modification in image editing evaluation, outperforming existing metrics.

Findings

AugCLIP correlates strongly with human evaluations across five datasets.

It outperforms existing metrics in diverse editing scenarios.

AugCLIP effectively models ideal edits in CLIP space.

Abstract

The development of vision-language and generative models has significantly advanced text-guided image editing, which seeks the preservation of core elements in the source image while implementing modifications based on the target text. However, existing metrics have a context-blindness problem, indiscriminately applying the same evaluation criteria on completely different pairs of source image and target text, biasing towards either modification or preservation. Directional CLIP similarity, the only metric that considers both source image and target text, is also biased towards modification aspects and attends to irrelevant editing regions of the image. We propose AugCLIP, a context-aware metric that adaptively coordinates preservation and modification aspects, depending on the specific context of a given source image and target text. This is done by deriving the CLIP representation of an ideally edited image, that preserves the source image with necessary modifications to align with target text. More specifically, using a multi-modal large language model, AugCLIP augments the textual descriptions of the source and target, then calculates a modification vector through a hyperplane that separates source and target attributes in CLIP space. Extensive experiments on five benchmark datasets, encompassing a diverse range of editing scenarios, show that AugCLIP aligns remarkably well with human evaluation standards, outperforming existing metrics. The code is available at https://github.com/augclip/augclip_eval.