Search papers, labs, and topics across Lattice.
This paper introduces iterative TARFlow (iTARFlow), a novel Normalizing Flow (NF) architecture that combines autoregressive generation with iterative denoising inspired by diffusion models. iTARFlow maintains a fully likelihood-based objective during training, unlike diffusion models. Experiments show iTARFlow achieves competitive performance on ImageNet at 64, 128, and 256 resolutions, demonstrating its potential as a strong generative model.
Normalizing Flows can now compete with diffusion models on image generation tasks, thanks to an iterative denoising scheme that boosts performance without sacrificing likelihood-based training.
Normalizing Flows (NFs) are a classical family of likelihood-based methods that have received revived attention. Recent efforts such as TARFlow have shown that NFs are capable of achieving promising performance on image modeling tasks, making them viable alternatives to other methods such as diffusion models. In this work, we further advance the state of Normalizing Flow generative models by introducing iterative TARFlow (iTARFlow). Unlike diffusion models, iTARFlow maintains a fully end-to-end, likelihood-based objective during training. During sampling, it performs autoregressive generation followed by an iterative denoising procedure inspired by diffusion-style methods. Through extensive experiments, we show that iTARFlow achieves competitive performance across ImageNet resolutions of 64, 128, and 256 pixels, demonstrating its potential as a strong generative model and advancing the frontier of Normalizing Flows. In addition, we analyze the characteristic artifacts produced by iTARFlow, offering insights that may shed light on future improvements. Code is available at https://github.com/apple/ml-itarflow.
