Counterfactual Image Editing
This work addresses the challenge of incorporating causal relationships into generative AI for image editing, which is incremental as it builds on existing methods by adding causal modeling.
The paper tackles the problem of counterfactual image editing by formalizing it with augmented structural causal models and proving two impossibility results regarding identifiability and guarantees. It proposes a relaxation using counterfactual-consistent estimators and develops an efficient algorithm based on neural causal models to generate images.
Counterfactual image editing is an important task in generative AI, which asks how an image would look if certain features were different. The current literature on the topic focuses primarily on changing individual features while remaining silent about the causal relationships between these features, as present in the real world. In this paper, we formalize the counterfactual image editing task using formal language, modeling the causal relationships between latent generative factors and images through a special type of model called augmented structural causal models (ASCMs). Second, we show two fundamental impossibility results: (1) counterfactual editing is impossible from i.i.d. image samples and their corresponding labels alone; (2) even when the causal relationships between the latent generative factors and images are available, no guarantees regarding the output of the model can be provided. Third, we propose a relaxation for this challenging problem by approximating non-identifiable counterfactual distributions with a new family of counterfactual-consistent estimators. This family exhibits the desirable property of preserving features that the user cares about across both factual and counterfactual worlds. Finally, we develop an efficient algorithm to generate counterfactual images by leveraging neural causal models.