AdaEraser: Training-Free Object Removal via Adaptive Attention Suppression

Object removal aims to eliminate specified objects from images while plausibly inpainting the affected regions with background content. Current training-free methods typically block attention to object regions within self-attention layers during the image generation process, leveraging surrounding background information to restore the image. However, indiscriminate suppression of self-attention in the vacated areas can degrade generation quality, as the model must simultaneously reconstruct background content in these regions. To solve this conflict, we propose AdaEraser, an adaptive framework that dynamically modulates attention based on the estimated presence of target object concepts. Through analysis of self-attention map evolution across denoising timesteps before and during removal, we develop a token-wise adaptive attention suppression strategy. This approach enables progressive perception of object removal throughout the denoising process, with the suppression strength in self-attention layers adjusted adaptively. Extensive experiments demonstrate that AdaEraser achieves superior performance in object removal, outperforming even training-based methods.

Removing unwanted objects from images is useful for photo editing, but it is difficult because the model must both erase the object and fill in the missing background naturally. Existing training-free methods often remove the object by forcing the image generation model to pay less attention to the object region. However, this can also weaken the model’s ability to reconstruct the background. We propose AdaEraser, a training-free object removal method that adaptively controls how much the model should reduce its attention to the object region during the image generation process. Instead of using a fixed suppression strength, AdaEraser estimates how much of the object still remains at each step and adjusts the suppression accordingly. This helps the model remove the target object while preserving useful background information. Our experiments show that AdaEraser produces cleaner removal results and more realistic backgrounds than previous methods, even compared with some methods that require extra training. This provides a simple and practical way to improve object removal using pretrained image generation models.