Fast method improves image editing with reward and content control
FASTER: Fast Adjoint Stochastic Transport for Endpoint Refinement in Reward-Guided Image Editing
Computer Vision and Pattern Recognition
Summary
Editing images using AI often tries to improve certain qualities while keeping the original content and style intact, but this can be slow because it relies on big, complex models repeatedly. The authors developed a new way to train small, fast networks that edit images by using feedback from both pre-trained models and reward signals, without needing to run the large model many times. Their approach, called FASTER, reuses information across updates to save time and resources. Tests show FASTER is significantly faster and performs better than previous methods on popular image editing models.
What this means in practice
- •For graphic design teams: Speed up AI-based image editing workflows that improve desired features while preserving original content using faster optimization methods.
- •For game developers: Enable quick reward-driven image adjustments during content creation pipelines without excessive computational costs from large generative models.
Authors
Yimiao Zhou, Zejia Zhong, Jingya Wang, Ye Shi
Abstract
Reward-guided image editing at test time seeks to improve a specified reward while preserving source content and visual plausibility. Many existing approaches optimize candidates through pretrained generation processes, making repeated adjustment depend on costly large-model execution and, in some cases, backbone backpropagation. We develop a theoretical framework that jointly accounts for reward, source preservation, and pretrained-prior preferences, allowing the desired output distribution to be specified separately from the dynamics used to realize it. Based on this framework, we introduce FASTER, which trains a small network for each source and objective to perform inexpensive editing, while pretrained and reward models provide feedback on candidate outputs. By reusing each candidate and its feedback across multiple small-network updates, FASTER reduces repeated sampling and supervision queries without placing the pretrained generative backbone inside the inner optimization loop. On SD3, FASTER leads all four target metrics and several validation metrics among the evaluated methods. Compared with the evaluated baseline that optimizes controls along pretrained generation trajectories, FASTER achieves editing-time speedups of up to \({6.91\times}\) on Stable Diffusion 3 and \({24.14\times}\) on Stable Diffusion 1.5.