Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

CycleNet: Rethinking Cycle Consistency in Text-Guided Diffusion for Image Manipulation

About

Diffusion models (DMs) have enabled breakthroughs in image synthesis tasks but lack an intuitive interface for consistent image-to-image (I2I) translation. Various methods have been explored to address this issue, including mask-based methods, attention-based methods, and image-conditioning. However, it remains a critical challenge to enable unpaired I2I translation with pre-trained DMs while maintaining satisfying consistency. This paper introduces Cyclenet, a novel but simple method that incorporates cycle consistency into DMs to regularize image manipulation. We validate Cyclenet on unpaired I2I tasks of different granularities. Besides the scene and object level translation, we additionally contribute a multi-domain I2I translation dataset to study the physical state changes of objects. Our empirical studies show that Cyclenet is superior in translation consistency and quality, and can generate high-quality images for out-of-domain distributions with a simple change of the textual prompt. Cyclenet is a practical framework, which is robust even with very limited training data (around 2k) and requires minimal computational resources (1 GPU) to train. Project homepage: https://cyclenetweb.github.io/

Sihan Xu, Ziqiao Ma, Yidong Huang, Honglak Lee, Joyce Chai• 2023

Related benchmarks

TaskDatasetResultRank
Time Series ForecastingETTh1
MSE0.432
869
Time Series ForecastingETTh2
MSE0.383
796
Time Series ForecastingETTm2
MSE0.266
552
Multivariate long-term forecastingETTh1
MSE0.432
509
Time Series ForecastingWeather
MSE0.243
497
Multivariate long-term series forecastingETTh2
MSE0.383
470
Multivariate long-term series forecastingWeather
MSE0.243
434
Multivariate long-term series forecastingETTm1
MSE0.379
404
Time Series ForecastingETTm1
MSE0.379
363
Time Series ForecastingECL
MAE0.259
345
Showing 10 of 32 rows

Other info

Code

Follow for update