Logo

D³-Predictor: Noise-Free Deterministic Diffusion for Dense Prediction

Changliang Xia*, Chengyou Jia*, Minnan Luo†, Zhuohang Dang, Xin Shen, Bowen Ping,
Xi'an Jiaotong University
*Equal Contribution. †Corresponding Author.
Paper arXiv Code Hugging Face Model Hugging Face Dataset
Logo

Demos and Comparison

Logo

Abstract

Although diffusion models with strong visual priors have emerged as powerful dense prediction backboens, they overlook a core limitation: the stochastic noise at the core of diffusion sampling is inherently misaligned with dense prediction that requires a deterministic mapping from image to geometry. In this paper, we show that this stochastic noise corrupts fine-grained spatial cues and pushes the model toward timestep-specific noise objectives, consequently destroying meaningful geometric structure mappings. To address this, we introduce D³-Predictor, a noise-free deterministic framework built by reformulating a pretrained diffusion model without stochasticity noise. Instead of relying on noisy inputs to leverage diffusion priors, D³-Predictor views the pretrained diffusion network as an ensemble of timestep-dependent visual experts and self-supervisedly aggregates their heterogeneous priors into a single, clean, and complete geometric prior. Meanwhile, we utilize task-specific supervision to seamlessly adapt this noise-free prior to dense prediction tasks. Extensive experiments on various dense prediction tasks demonstrate that D³-Predictor achieves competitive or state-of-the-art performance in diverse scenarios. In addition, it requires less than half the training data previously used and efficiently performs inference in a single step.

Logo

D³-Predictor Architecture

ap-clip

Overview of the D³-Predictor. (a) We reformulate the pretrained diffusion model into a noise-free framework to better suit dense prediction tasks, without compromising the diffusion prior with minimal overhead. (b) The D³-Predictor takes a clean image as input and produces an accurate prediction with impressive geometric details in a single step.

BibTeX

If you find our paper or code useful for your research, please consider citing our work. 

@misc{xia2025mathrmdmathrm3predictornoisefreedeterministicdiffusion,
          title={$\mathrm{D}^{\mathrm{3}}$-Predictor: Noise-Free Deterministic Diffusion for Dense Prediction}, 
          author={Changliang Xia and Chengyou Jia and Minnan Luo and Zhuohang Dang and Xin Shen and Bowen Ping},
          year={2025},
          eprint={2512.07062},
          archivePrefix={arXiv},
          primaryClass={cs.CV},
          url={https://arxiv.org/abs/2512.07062}, 
        }