Binici, K., Acar, C., Aggarwal, S., Liu, S., & Mitra, T. (2025). Efficient Text-to-Image Generation: An Adaptive Step Schedule Controller for Diffusion Models. 2025 IEEE International Conference on Image Processing (ICIP), 355–360. https://doi.org/10.1109/icip55913.2025.11084691
Abstract:
Text-to-image diffusion models often use a fixed number
of denoising steps, balancing time costs and image qual-
ity. However, the optimal number of steps depends on the
complexity of the input text prompt. We propose an adap-
tive diffusion controller that dynamically adjusts the number
of steps to generate high-quality images efficiently, without
additional model training. By leveraging a mixture of step
schedules with varying step sizes and evaluating the error
term discrepancy at each timestep, our method transitions
between schedules to optimize performance. Experiments
on COCO and DiffusionDB show that our approach reduces
inference time while maintaining visual fidelity, offering a
more efficient alternative for text-to-image diffusion models.
License type:
Publisher Copyright
Funding Info:
Student was funded by A*STAR SINGA PhD scholarship