RF-MatID: Dataset and Benchmark for Radio Frequency Material Identification

TL;DR

RF-MatID provides the first open, large-scale RF dataset for fine-grained material identification, spanning $4$ to $43.5$ GHz with 16 materials organized into 5 superclasses and totaling 142k samples across frequency- and time-domain representations. It introduces geometry perturbations (incidence angle and stand-off distance) and a comprehensive multi-protocol benchmark to evaluate in- and out-of-distribution robustness, using both raw frequency-domain data and time-domain transforms. The work benchmarks nine models and a robust baseline across five frequency protocols and seven data splits, revealing the practical viability of frequency-domain RF signals for material discrimination and highlighting domain-shift challenges and opportunities for physical-constraint and domain-generalization strategies. By providing an open, standardized dataset and benchmarking framework, RF-MatID aims to accelerate reproducible research, foster cross-domain robustness, and advance RF-based material identification for indoor embodied AI applications.

Abstract

Accurate material identification plays a crucial role in embodied AI systems, enabling a wide range of applications. However, current vision-based solutions are limited by the inherent constraints of optical sensors, while radio-frequency (RF) approaches, which can reveal intrinsic material properties, have received growing attention. Despite this progress, RF-based material identification remains hindered by the lack of large-scale public datasets and the limited benchmarking of learning-based approaches. In this work, we present RF-MatID, the first open-source, large-scale, wide-band, and geometry-diverse RF dataset for fine-grained material identification. RF-MatID includes 16 fine-grained categories grouped into 5 superclasses, spanning a broad frequency range from 4 to 43.5 GHz, and comprises 142k samples in both frequency- and time-domain representations. The dataset systematically incorporates controlled geometry perturbations, including variations in incidence angle and stand-off distance. We further establish a multi-setting, multi-protocol benchmark by evaluating state-of-the-art deep learning models, assessing both in-distribution performance and out-of-distribution robustness under cross-angle and cross-distance shifts. The 5 frequency-allocation protocols enable systematic frequency- and region-level analysis, thereby facilitating real-world deployment. RF-MatID aims to enable reproducible research, accelerate algorithmic advancement, foster cross-domain robustness, and support the development of real-world application in RF-based material identification.

Figures (8)

• Figure 1: Data collection setup: (a) the customized sensing platform and (b) the acquisition layout.
• Figure 2: The visual illustration of the 16 fine-grained material categories.
• Figure 3: Preliminary experiments on consecutive sub-bands with various bandwidths. Accuracy in percentage value is applied for performance evaluation.
• Figure 4: Comparison between material identification performance on time- and frequency-domain signals.
• Figure 4: The visual band allocation of each frequency protocol. The frequency values are in GHz.