Abstract:3D human reconstruction from RGB images achieves decent results in good weather conditions but degrades dramatically in rough weather. Complementarily, mmWave radars have been employed to reconstruct 3D human joints and meshes in rough weather. However, combining RGB and mmWave signals for weather-robust 3D human reconstruction is still an open challenge, given the sparse nature of mmWave and the vulnerability of RGB images. The limited research about the impact of missing points and sparsity features of mmWave data on reconstruction performance, as well as the lack of available datasets for paired mmWave-RGB data, further complicates the process of fusing the two modalities. To fill these gaps, we build up an automatic 3D body annotation system with multiple sensors to collect a large-scale mmWave dataset. The dataset consists of synchronized and calibrated mmWave radar point clouds and RGB(D) images under different weather conditions and skeleton/mesh annotations for humans in these scenes. With this dataset, we conduct a comprehensive analysis about the limitations of single-modality reconstruction and the impact of missing points and sparsity on the reconstruction performance. Based on the guidance of this analysis, we design ImmFusion, the first mmWave-RGB fusion solution to robustly reconstruct 3D human bodies in various weather conditions. Specifically, our ImmFusion consists of image and point backbones for token feature extraction and a Transformer module for token fusion. The image and point backbones refine global and local features from original data, and the Fusion Transformer Module aims for effective information fusion of two modalities by dynamically selecting informative tokens. Extensive experiments demonstrate that ImmFusion can efficiently utilize the information of two modalities to achieve robust 3D human body reconstruction in various weather environments. In addition, our method achieves superior accuracy compared to that of the state-of-the-art Transformer-based LiDAR-camera fusion methods.

Microwave Imaging of Human Body from Point Cloud with Capsule Network

High Resolution 3D Microwave Real Time Imaging System Based on Intelligent Metasurface

Mmskeleton: 3D Human Skeleton Estimation Using Millimeter Wave Radar Sparse Point Clouds

Towards Weather-Robust 3D Human Body Reconstruction: Millimeter-Wave Radar-Based Dataset, Benchmark, and Multi-Modal Fusion

Microwave Imaging of Non-Rigid Moving Target Using 2D Sparse MIMO Array

1-D Microwave Imaging of Human Cardiac Motion: An Ab-Initio Investigation

Microwave Imaging Customized on Demand under Random Field Illumination

3D Microwave Imaging As Point Cloud with Continous Normalizing Flows

High-resolution Three-dimensional Microwave Imaging Using a Generative Adversarial Network

Microwave Imaging under Oblique Illumination

Metamaterials Based Intelligent Microwave Human Behavior Recognition

Machine-learning Reprogrammable Metasurface Imager.

Real-time dangerous objects detection in millimeter wave images

Intelligent Metasurface Imager and Recognizer

Microwave Reconstruction of 3D Human Facial Landmarks Using a Programmable Metasurface

Microwave Imaging and Classification of Hand Gestures

Visual Intrusion Detection Based On CBAM-Capsule Networks

mmID: High-Resolution mmWave Imaging for Human Identification

Experimental Setup with a Compact Vector Network Analyzer, RGB-D Sensor, and Linear Drive Modules for Imitation Modeling of Microwave Personnel Screening System in Motion

Optimal Imaging and Semantic Segmentation of the Human Body for Passive THz Security Inspection

Real-Time Detection of Concealed Threats with Passive Millimeter Wave and Visible Images via Deep Neural Networks