Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Talk2Radar: Bridging Natural Language with 4D mmWave Radar for 3D Referring Expression Comprehension

Runwei Guan, Ruixiao Zhang, Ningwei Ouyang, Jianan Liu, Ka Lok Man, Xiaohao Cai, Ming Xu, Jeremy Smith, Eng Gee Lim, Yutao Yue, Hui Xiong

TL;DR

Talk2Radar addresses radar-based 3D visual grounding by introducing Talk2Radar, the first dataset for 4D mmWave radar–driven 3D REC, and a model, T-RadarNet, that fuses radar point clouds with natural language prompts. The method combines Gated Graph Fusion for cross-modal alignment and Deformable-FPN to model irregular radar point clouds, achieving state-of-the-art results on Talk2Radar. The dataset contains 8,682 prompts and 20,558 referent objects, with prompts capturing both qualitative and quantitative radar-perceivable attributes and allowing multi-object references. This work enables robust, all-weather, language-driven object localization for autonomous driving.

Abstract

Embodied perception is essential for intelligent vehicles and robots in interactive environmental understanding. However, these advancements primarily focus on vision, with limited attention given to using 3D modeling sensors, restricting a comprehensive understanding of objects in response to prompts containing qualitative and quantitative queries. Recently, as a promising automotive sensor with affordable cost, 4D millimeter-wave radars provide denser point clouds than conventional radars and perceive both semantic and physical characteristics of objects, thereby enhancing the reliability of perception systems. To foster the development of natural language-driven context understanding in radar scenes for 3D visual grounding, we construct the first dataset, Talk2Radar, which bridges these two modalities for 3D Referring Expression Comprehension (REC). Talk2Radar contains 8,682 referring prompt samples with 20,558 referred objects. Moreover, we propose a novel model, T-RadarNet, for 3D REC on point clouds, achieving State-Of-The-Art (SOTA) performance on the Talk2Radar dataset compared to counterparts. Deformable-FPN and Gated Graph Fusion are meticulously designed for efficient point cloud feature modeling and cross-modal fusion between radar and text features, respectively. Comprehensive experiments provide deep insights into radar-based 3D REC. We release our project at https://github.com/GuanRunwei/Talk2Radar.

Talk2Radar: Bridging Natural Language with 4D mmWave Radar for 3D Referring Expression Comprehension

TL;DR

Talk2Radar addresses radar-based 3D visual grounding by introducing Talk2Radar, the first dataset for 4D mmWave radar–driven 3D REC, and a model, T-RadarNet, that fuses radar point clouds with natural language prompts. The method combines Gated Graph Fusion for cross-modal alignment and Deformable-FPN to model irregular radar point clouds, achieving state-of-the-art results on Talk2Radar. The dataset contains 8,682 prompts and 20,558 referent objects, with prompts capturing both qualitative and quantitative radar-perceivable attributes and allowing multi-object references. This work enables robust, all-weather, language-driven object localization for autonomous driving.

Abstract

Embodied perception is essential for intelligent vehicles and robots in interactive environmental understanding. However, these advancements primarily focus on vision, with limited attention given to using 3D modeling sensors, restricting a comprehensive understanding of objects in response to prompts containing qualitative and quantitative queries. Recently, as a promising automotive sensor with affordable cost, 4D millimeter-wave radars provide denser point clouds than conventional radars and perceive both semantic and physical characteristics of objects, thereby enhancing the reliability of perception systems. To foster the development of natural language-driven context understanding in radar scenes for 3D visual grounding, we construct the first dataset, Talk2Radar, which bridges these two modalities for 3D Referring Expression Comprehension (REC). Talk2Radar contains 8,682 referring prompt samples with 20,558 referred objects. Moreover, we propose a novel model, T-RadarNet, for 3D REC on point clouds, achieving State-Of-The-Art (SOTA) performance on the Talk2Radar dataset compared to counterparts. Deformable-FPN and Gated Graph Fusion are meticulously designed for efficient point cloud feature modeling and cross-modal fusion between radar and text features, respectively. Comprehensive experiments provide deep insights into radar-based 3D REC. We release our project at https://github.com/GuanRunwei/Talk2Radar.
Paper Structure (19 sections, 6 equations, 5 figures, 7 tables)

This paper contains 19 sections, 6 equations, 5 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. 3D Referring Expression Comprehension in Traffic
  4. 3D Object Detection with Point Cloud
  5. Talk2Radar Dataset
  6. Data Collection and Annotation Process
  7. Dataset Statistics
  8. Metrics and Subset Settings
  9. Methods
  10. Overall Pipeline
  11. Gated Graph Fusion
  12. Deformable FPN
  13. Training Objectives
  14. Experiments
  15. Implementation Settings
  16. ...and 4 more sections

Figures (5)

  • Figure 1: Samples in Talk2Radar. The first and second row respectively presents referred objects by point cloud sensors (4D radar and LiDAR) and camera.
  • Figure 2: Annotation process of Talk2Radar dataset.
  • Figure 3: Statictics of Talk2Radar dataset on referent objects and prompts.
  • Figure 4: The architecture of T-RadarNet. LiDAR is not the main modality, but it can also be used as the input of T-RadarNet.
  • Figure 5: Prediction by T-RadarNet. The first row presents correct cases while the second shows problematic cases (FN: False Negative, FP: False Positive).