RadarSFD: Single-Frame Diffusion with Pretrained Priors for Radar Point Clouds
Bin Zhao, Nakul Garg
AI summary
Problem
Current radar imaging requires multi-frame aggregation or synthetic aperture to improve resolution, which is impractical for compact, SWaP-constrained robotic platforms. The paper asks whether a single radar capture can yield a dense, LiDAR-like point cloud suitable for real-time perception.
Approach
The method conditions a latent diffusion model, initialized with pretrained monocular depth priors, on raw single-frame radar BEV images using channel-wise latent concatenation and a dual-space loss to enforce structural fidelity.
Key results
- Achieves state-of-the-art single-frame radar-to-LiDAR translation on the RadarHD benchmark
- Reduces Chamfer distance from 56 cm to 35 cm and Modified Hausdorff distance from 45 cm to 28 cm versus single-frame baselines
- Recovers fine geometric structures like narrow gaps and walls without sacrificing native 4 cm range resolution
- Demonstrates strong generalization across unseen indoor and outdoor environments
Why it matters
Enables high-resolution, real-time spatial perception for size-, weight-, and power-constrained robots and drones operating in challenging environments like fog or low light.
Abstract
Millimeter-wave radar provides perception robust to fog, smoke, dust, and low light, making it attractive for size, weight, and power constrained robotic platforms. Current radar imaging methods, however, rely on synthetic aperture or multi-frame aggregation to improve resolution, which is impractical for small aerial, inspection, or wearable systems. We present RadarSFD, a conditional latent diffusion framework that reconstructs dense LiDAR-like point clouds from a single radar frame without motion or SAR. Our approach transfers geometric priors from a pretrained monocular depth estimator into the diffusion backbone, anchors them to radar inputs via channel-wise latent concatenation, and regularizes outputs with a dual-space objective combining latent and pixel-space losses. On the RadarHD benchmark, RadarSFD achieves state-of-the- art performance against baseline models. Qualitative results show recovery of fine walls and narrow gaps, and experiments across new environments confirm strong generalization. Abla- tion studies highlight the importance of pretrained initialization, radar BEV conditioning, and the dual-space loss. Together, these results establish the first practical single-frame, no-SAR mmWave radar pipeline for dense point cloud perception in compact robotic systems. The project page is available at https://phi-lab-rice.github.io/RadarSFD/