Many paired image-to-image translation tasks admit multiple valid target appearances, while training provides only one target per source. We propose One-to-Many Flow (O2MFlow), a latent-conditioned Rectified Flow (RF) model for one-to-many generation and backward translation (target→source) within a single transport framework. Conditioning the RF dynamics on an appearance latent turns transport into a family of latent-indexed bijections: a fixed latent gives a mapping that is inverted by time-reversed integration, and sampled latents produce diverse outputs for the same input. We formalize reusable appearance variations as Shared Latent Transport and show that Kullback--Leibler (KL) regularization directly controls the source--latent distribution gap between training and generation. Across three benchmarks, O2MFlow offers dataset-dependent forward diversity--fidelity trade-offs and the lowest backward LPIPS-Dist among evaluated methods, with the lowest backward FID on two benchmarks.
Figures & tables
Figure 1: Four representative settings of image-to-image translation. The proposed method corresponds to (d), supporting one-to-many paired translation and backward translation within a single framework.
Figure 2: (a) Standard RF. (b) O2MFlow samples appearance latents for diverse forward translation and infers a latent from the target for backward translation. Here q denotes the target-only encoder and zX1 a latent inferred from target X1 ; parameter subscripts are omitted. The vertical axis represents image space.
Figure 3: Shared Latent Transport on edges2shoes. Reusing a target-encoded latent across sources produces related appearance tendencies while preserving source structure.
Figure 4: Training overview. RF regression fits paired target endpoints, and KL regularization aligns the target-encoded latent distribution with the prior. Flame symbols mark the jointly trained encoder and velocity field.
Figure 5: Same-latent round trip; the same fixed z is used in both directions.
Figure 6: Qualitative comparison of forward translation on paired translation datasets.
Figure 7: Forward translation results obtained by sampling different appearance latents.
Figure 8: Comparison of output diversity on the facades dataset.
Figure 9: Backward translation results on paired translation datasets. We compare our single-model backward translation against pix2pix trained separately for the backward direction, CycleGAN, SymmFlow, and BDBM.
Figure 10: Forward diversity–fidelity trade-off on three datasets. The upper-left region combines higher LPIPS-Div with lower LPIPS-Dist.
Dataset /Method
LPIPS-Div ↑
LPIPS-Dist ↓
FID ↓
edges2shoes
pix2pix
–
0.1630
99.71
CycleGAN
–
0.2050
108.88
BicycleGAN
0.1914
0.2130
58.51
DRIT++
0.1954
0.2280
50.18
BBDM
0.1089
0.1280
37.44
Table 1: Quantitative comparison on forward translation.
Dataset /Method
LPIPS-Dist ↓
FID ↓
edges2shoes
pix2pix (baseline)
0.0501
36.68
CycleGAN
0.2309
197.79
SymmFlow
0.0834
109.98
BDBM
0.0702
43.97
O2MFlow
0.0287
22.57
Table 2: Quantitative comparison on backward translation.
Figure 11: Latent interpolation: appearance changes smoothly while the source content is preserved.
Components
Forward
Backward
Latent
KL
LPIPS-Div ↑
LPIPS-Dist ↓
FID ↓
LPIPS-Dist ↓
FID ↓
edges2shoes
–
–
–
0.1013
54.02
0.0277
23.09
✓
–
0.3269
0.3187
83.28
0.0281
23.36
✓
✓
0.1776
0.1924
50.86
0.0287
22.57
edges2handbags
Table 3: Ablation study of the proposed components on the edges2shoes, edges2handbags, and facades datasets. Latent denotes appearance-latent conditioning; KL denotes regularization of the target-encoded latent distribution toward the prior by LKL in Eq. ( 7 ). Checkmarks indicate enabled components.
Figure 12: Effect of KL regularization on edges2shoes. The top and bottom rows show samples without and with KL, respectively; Table 3 reports the quantitative ablation.
Bidirectional unpaired image translation must preserve source-specific structure while learning coherent transformations in both directions without paired supervision. Existing methods typically employ two direction-specific generators or train separate one-way models. Even when linked by cycle consistency, such models constrain only the round-trip endpoint reconstruction, without requiring the two directions to obey a common local transformation rule. We propose UniCycleFlow, a rectified-flow framework that represents bidirectional translation as forward and reverse integration of a single time-conditioned velocity field. This formulation organizes both directions within the same continuous dynamics, rather than coupling otherwise separate endpoint mappings. A key challenge is that unpaired data provide no meaningful source--target coupling from which rectified-flow trajectories can be constructed. UniCycleFlow addresses this challenge by learning deterministic source-conditioned endpoints whose marginal distributions are adversarially matched to the opposite domains. The resulting paths are regularized by stop-gradient self-flow matching for intermediate velocity supervision, discrete cycle closure for forward--reverse consistency, and representation path-velocity regularization for controlling localized feature changes along the trajectory. Across ten translation directions, UniCycleFlow achieves the lowest FID on 7 of 10 tasks using a single Euler evaluation and obtains the best average FID of 55.1.
Xianhao Zhou, Jianghao Wu, Shaoting Zhang +1
School of Mechanical and Electrical Engineering, University of Electronic Science and Technology of China, Chengdu, China · Faculty of Information Technology, Monash University, Melbourne, Australia · Shanghai Artificial Intelligence Laboratory, Shanghai, China
In this work, we propose Image-to-Image Rectified Flow Reformulation (I2I-RFR), a practical plug-in reformulation that recasts standard I2I regression networks as continuous-time transport models. While pixel-wise I2I regression is simple, stable, and easy to adapt across tasks, it often over-smooths ill-posed and multimodal targets, whereas generative alternatives often require additional components, task-specific tuning, and more complex training and inference pipelines. Our method augments the backbone input by channel-wise concatenation with a noise-corrupted version of the ground-truth target and optimizes a simple t-reweighted pixel loss. This objective admits a rectified-flow interpretation via an induced velocity field, enabling ODE-based progressive refinement at inference time while largely preserving the standard supervised training pipeline. In most cases, adopting I2I-RFR requires only expanding the input channels, and inference can be performed with a few explicit solver steps (e.g., 3 steps) without distillation. Extensive experiments across multiple image-to-image translation and video restoration tasks show that I2I-RFR generally improves performance across a wide range of tasks and backbones, with particularly clear gains in perceptual quality and detail preservation. Overall, I2I-RFR provides a lightweight way to incorporate continuous-time refinement into conventional I2I models without requiring a heavy generative pipeline.
In-image machine translation (IIMT) replaces text in an image with its translation while preserving the surrounding visual context. Existing task-specific models achieve convincing performance but require massive paired training images before and after translation. As advanced text-to-image (T2I) models have reached a practically usable level of multilingual text generation, we argue that this capability can be transferred to IIMT and present FlowTranslator. It is a training-free IIMT framework that first employs an OCR model and an LLM to generate the translation prompt, which is injected into the T2I-based diffusion model. To optimize the accurate and natural generation of translated text, we introduce two trajectory controls: Region Coupling Release (RCR) partially decorrelates target-branch noise inside the text region early in the trajectory, while Coupling Expansion (CE) amplifies the emerging edit displacement mid-trajectory. Under the EasyOCR read-back protocol, FlowTranslator achieves average PRIM BLEU and COMET scores of 20.6 and 57.2, respectively, substantially outperforming the strong training-based baseline UniTranslator. On the more challenging and realistic whole-image ScenePair photographs translated into Chinese, it reaches 40.1% exact-match accuracy and surpasses existing editing systems, some of which only address the text-rendering stage rather than the complete IIMT task. These results establish flow-based editing as a strong, fully training-free baseline for IIMT.
Zhixian Li, Xingsong Ye, Yongkun Du +6
Institute of Trustworthy Embodied AI, Fudan University · Shanghai Key Laboratory of Multimodal Embodied AI