Hao Tang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Hao Tang has a total of 53 co-authors. Here are the top co-authors by number of shared publications:
- Zhigang Zhu
Department of Computer Science, The Graduate Center—CUNY, New York, NY 10016, USA
11 shared publications - Jiawei Liu
The Graduate Center of The City University of New York
3 shared publications - Oyewole Oyekoya
Hunter College, City University of New York, New York, NY, USA
3 shared publications - Gavin Gong
East Brunswick High School
2 shared publications - Hong Zhao
Department of Computer Science, The City College of New York, New York, New York, USA
2 shared publications - Jiahao Xia
Department of Civil and Environmental Engineering Rutgers The State University of New Jersey
2 shared publications
Latest Publications
- Open Access
A Single-Pass Approach That Mines Unstructured Robotic Trajectories for Calibrating Surveillance Cameras
The ubiquity of surveillance networks in public infrastructure presents a significant, yet underutilized, opportunity to assist vulnerable populations, particularly individuals who are blind or have low vision (BLV). However, transforming these passive video feeds into active guidance systems requires accurate camera calibration, a process that is traditionally labor-intensive and unscalable in large facilities. This paper introduces a novel, automated framework that leverages a mobile quadruped robot (Boston Dynamics Spot) as a dynamic calibration agent. We propose a single-pass approach with two planar calibration algorithms, which mines unstructured robotic trajectories to construct distinct geometric features on the ground plane. By synthesizing “virtual rectangles” from the robot’s odometry, our method first recovers camera focal length through vanishing point estimation by constructing virtual rectangles, and then solves for extrinsic 6-DoF pose using two algorithms: our proposed Virtual Rectangle (ViR) algorithm and a standard planar Perspective-n-Point (PnP) algorithm. Experimental validation using real-world data demonstrates the system’s robustness to sensor noise, maintaining focal length errors around , rotational errors around and relative translation errors of approximately , while synthetic simulations indicate focal length errors generally under 10%, rotational errors under , and translation errors also around 10% despite heavy image and object disturbances. This work eliminates the need for manual calibration targets for dynamic camera calibration, effectively converting static security infrastructure into a metric sensing network capable of supporting high-fidelity social robotics applications.Open AccessA Single-Pass Approach That Mines Unstructured Robotic Trajectories for Calibrating Surveillance Cameras
The ubiquity of surveillance networks in public infrastructure presents a significant, yet underutilized, opportunity to assist vulnerable populations, particularly individuals who are blind or have low vision (BLV). However, transforming these passive video feeds into active guidance systems requires accurate camera calibration, a process that is traditionally labor-intensive and unscalable in large facilities. This paper introduces a novel, automated framework that leverages a mobile quadruped robot (Boston Dynamics Spot) as a dynamic calibration agent. We propose a single-pass approach with two planar calibration algorithms, which mines unstructured robotic trajectories to construct distinct geometric features on the ground plane. By synthesizing “virtual rectangles” from the robot’s odometry, our method first recovers camera focal length through vanishing point estimation by constructing virtual rectangles, and then solves for extrinsic 6-DoF pose using two algorithms: our proposed Virtual Rectangle (ViR) algorithm and a standard planar Perspective-n-Point (PnP) algorithm. Experimental validation using real-world data demonstrates the system’s robustness to sensor noise, maintaining focal length errors around , rotational errors around and relative translation errors of approximately , while synthetic simulations indicate focal length errors generally under 10%, rotational errors under , and translation errors also around 10% despite heavy image and object disturbances. This work eliminates the need for manual calibration targets for dynamic camera calibration, effectively converting static security infrastructure into a metric sensing network capable of supporting high-fidelity social robotics applications. - Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-14
From Selfie Stick to Virtual Cane: Enabling Blind Exploration through Mobile Virtual Reality
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-14
Open AccessFrom Selfie Stick to Virtual Cane: Enabling Blind Exploration through Mobile Virtual Reality
- Open Access
- 3 December 2025
- Published by MDPI AG in Journal of Imaging
- Vol. 11 (12) , 430
Editorial on the Special Issue “Image and Video Processing for Blind and Visually Impaired”
- 3 December 2025
- Published by MDPI AG in Journal of Imaging
- Vol. 11 (12) , 430
Open AccessEditorial on the Special Issue “Image and Video Processing for Blind and Visually Impaired”
- Open Access
- 27 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 83-87
Enhancing Virtual Mobility for Individuals Who Are Blind or Have Low Vision: A Stationary Exploration Method
Designing accessible locomotion methods for individuals who are blind or have low vision (BLV) is a complex challenge, particularly in mobile VR environments with limited interface options. In this paper, we propose a novel locomotion technique on mobile VR that enables users to control a virtual character’s movement while staying stationary or within a small physical area. The technique utilizes the phone’s gyroscope for movement control, while providing spatial audio and vibration feedback to enhance virtual exploration for BLV individuals. Our study examines how BLV individuals acquire spatial knowledge in mobile VR environments. A user study is conducted to assess the effectiveness of the proposed approach.
- 27 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 83-87
Open AccessEnhancing Virtual Mobility for Individuals Who Are Blind or Have Low Vision: A Stationary Exploration Method
Designing accessible locomotion methods for individuals who are blind or have low vision (BLV) is a complex challenge, particularly in mobile VR environments with limited interface options. In this paper, we propose a novel locomotion technique on mobile VR that enables users to control a virtual character’s movement while staying stationary or within a small physical area. The technique utilizes the phone’s gyroscope for movement control, while providing spatial audio and vibration feedback to enhance virtual exploration for BLV individuals. Our study examines how BLV individuals acquire spatial knowledge in mobile VR environments. A user study is conducted to assess the effectiveness of the proposed approach.
- Open Access
VerifierQ: Enhancing LLM Test Time Compute with Q-Learning-based Verifiers
Recent advancements in test time compute, particularly through the use of verifier models, have significantly enhanced the reasoning capabilities of Large Language Models (LLMs). This generator-verifier approach closely resembles the actor-critic framework in reinforcement learning (RL). However, current verifier models in LLMs often rely on supervised fine-tuning without temporal difference learning such as Q-learning. This paper introduces VerifierQ, a novel approach that integrates Offline Q-learning into LLM verifier models. We address three key challenges in applying Q-learning to LLMs: (1) handling utterance-level Markov Decision Processes (MDPs), (2) managing large action spaces, and (3) mitigating overestimation bias. VerifierQ introduces a modified Bellman update for bounded Q-values, incorporates Implicit Q-learning (IQL) for efficient action space management, and integrates a novel Conservative Q-learning (CQL) formulation for balanced Q-value estimation. Our method enables parallel Q-value computation and improving training efficiency. While recent work has explored RL techniques like MCTS for generators, VerifierQ is among the first to investigate the verifier (critic) aspect in LLMs through Q-learning. This integration of RL principles into verifier models complements existing advancements in generator techniques, potentially enabling more robust and adaptive reasoning in LLMs. Experimental results on mathematical reasoning tasks demonstrate VerifierQ's superior performance compared to traditional supervised fine-tuning approaches, with improvements in efficiency, accuracy and robustness. By enhancing the synergy between generation and evaluation capabilities, VerifierQ contributes to the ongoing evolution of AI systems in addressing complex cognitive tasks across various domains.
Open AccessVerifierQ: Enhancing LLM Test Time Compute with Q-Learning-based Verifiers
Recent advancements in test time compute, particularly through the use of verifier models, have significantly enhanced the reasoning capabilities of Large Language Models (LLMs). This generator-verifier approach closely resembles the actor-critic framework in reinforcement learning (RL). However, current verifier models in LLMs often rely on supervised fine-tuning without temporal difference learning such as Q-learning. This paper introduces VerifierQ, a novel approach that integrates Offline Q-learning into LLM verifier models. We address three key challenges in applying Q-learning to LLMs: (1) handling utterance-level Markov Decision Processes (MDPs), (2) managing large action spaces, and (3) mitigating overestimation bias. VerifierQ introduces a modified Bellman update for bounded Q-values, incorporates Implicit Q-learning (IQL) for efficient action space management, and integrates a novel Conservative Q-learning (CQL) formulation for balanced Q-value estimation. Our method enables parallel Q-value computation and improving training efficiency. While recent work has explored RL techniques like MCTS for generators, VerifierQ is among the first to investigate the verifier (critic) aspect in LLMs through Q-learning. This integration of RL principles into verifier models complements existing advancements in generator techniques, potentially enabling more robust and adaptive reasoning in LLMs. Experimental results on mathematical reasoning tasks demonstrate VerifierQ's superior performance compared to traditional supervised fine-tuning approaches, with improvements in efficiency, accuracy and robustness. By enhancing the synergy between generation and evaluation capabilities, VerifierQ contributes to the ongoing evolution of AI systems in addressing complex cognitive tasks across various domains.
- Open Access
- 4 April 2024
- Published by MDPI AG in Information
- Vol. 15 (4) , 201
Flying Projectile Attitude Determination from Ground-Based Monocular Imagery with a Priori Knowledge
This paper discusses using ground-based imagery to determine the attitude of a flying projectile assuming prior knowledge of its external geometry. It presents a segmentation-based approach to follow the object and evaluates it quantitatively with simulated data and qualitatively with both simulated and real data. Two experimental cases are considered: One assumes reliable target distance measurement from an auxiliary range sensor, while the other assumes no range information. The results show that in the case of an unknown projectile–camera distance, with projectile dimensions of 1.378 m and 0.08 m in length and diameter, the estimated distance, in-plane location, and pitch angle accuracies are about 50 m, 0.15 m, and 6 degrees, respectively. Yaw angle estimation is ambiguous. In the second case, assuming that the projectile–camera distance is known resolves the ambiguity of yaw estimation, resulting in accuracies of about 0.15 m, 3 degrees, and 20 degrees for in-plane location, pitch, and yaw angles, respectively. These accuracies were normalized to a 1-km projectile–camera distance.
- 4 April 2024
- Published by MDPI AG in Information
- Vol. 15 (4) , 201
Open AccessFlying Projectile Attitude Determination from Ground-Based Monocular Imagery with a Priori Knowledge
This paper discusses using ground-based imagery to determine the attitude of a flying projectile assuming prior knowledge of its external geometry. It presents a segmentation-based approach to follow the object and evaluates it quantitatively with simulated data and qualitatively with both simulated and real data. Two experimental cases are considered: One assumes reliable target distance measurement from an auxiliary range sensor, while the other assumes no range information. The results show that in the case of an unknown projectile–camera distance, with projectile dimensions of 1.378 m and 0.08 m in length and diameter, the estimated distance, in-plane location, and pitch angle accuracies are about 50 m, 0.15 m, and 6 degrees, respectively. Yaw angle estimation is ambiguous. In the second case, assuming that the projectile–camera distance is known resolves the ambiguity of yaw estimation, resulting in accuracies of about 0.15 m, 3 degrees, and 20 degrees for in-plane location, pitch, and yaw angles, respectively. These accuracies were normalized to a 1-km projectile–camera distance.
- Open Access
- 21 February 2024
- Published by MDPI AG in Journal of Imaging
- Vol. 10 (3) , 52
Pedestrian-Accessible Infrastructure Inventory: Enabling and Assessing Zero-Shot Segmentation on Multi-Mode Geospatial Data for All Pedestrian Types
In this paper, a Segment Anything Model (SAM)-based pedestrian infrastructure segmentation workflow is designed and optimized, which is capable of efficiently processing multi-sourced geospatial data, including LiDAR data and satellite imagery data. We used an expanded definition of pedestrian infrastructure inventory, which goes beyond the traditional transportation elements to include street furniture objects that are important for accessibility but are often omitted from the traditional definition. Our contributions lie in producing the necessary knowledge to answer the following three questions. First, how can mobile LiDAR technology be leveraged to produce comprehensive pedestrian-accessible infrastructure inventory? Second, which data representation can facilitate zero-shot segmentation of infrastructure objects with SAM? Third, how well does the SAM-based method perform on segmenting pedestrian infrastructure objects? Our proposed method is designed to efficiently create pedestrian-accessible infrastructure inventory through the zero-shot segmentation of multi-sourced geospatial datasets. Through addressing three research questions, we show how the multi-mode data should be prepared, what data representation works best for what asset features, and how SAM performs on these data presentations. Our findings indicate that street-view images generated from mobile LiDAR point-cloud data, when paired with satellite imagery data, can work efficiently with SAM to create a scalable pedestrian infrastructure inventory approach with immediate benefits to GIS professionals, city managers, transportation owners, and walkers, especially those with travel-limiting disabilities, such as individuals who are blind, have low vision, or experience mobility disabilities.
- 21 February 2024
- Published by MDPI AG in Journal of Imaging
- Vol. 10 (3) , 52
Open AccessPedestrian-Accessible Infrastructure Inventory: Enabling and Assessing Zero-Shot Segmentation on Multi-Mode Geospatial Data for All Pedestrian Types
In this paper, a Segment Anything Model (SAM)-based pedestrian infrastructure segmentation workflow is designed and optimized, which is capable of efficiently processing multi-sourced geospatial data, including LiDAR data and satellite imagery data. We used an expanded definition of pedestrian infrastructure inventory, which goes beyond the traditional transportation elements to include street furniture objects that are important for accessibility but are often omitted from the traditional definition. Our contributions lie in producing the necessary knowledge to answer the following three questions. First, how can mobile LiDAR technology be leveraged to produce comprehensive pedestrian-accessible infrastructure inventory? Second, which data representation can facilitate zero-shot segmentation of infrastructure objects with SAM? Third, how well does the SAM-based method perform on segmenting pedestrian infrastructure objects? Our proposed method is designed to efficiently create pedestrian-accessible infrastructure inventory through the zero-shot segmentation of multi-sourced geospatial datasets. Through addressing three research questions, we show how the multi-mode data should be prepared, what data representation works best for what asset features, and how SAM performs on these data presentations. Our findings indicate that street-view images generated from mobile LiDAR point-cloud data, when paired with satellite imagery data, can work efficiently with SAM to create a scalable pedestrian infrastructure inventory approach with immediate benefits to GIS professionals, city managers, transportation owners, and walkers, especially those with travel-limiting disabilities, such as individuals who are blind, have low vision, or experience mobility disabilities.
- Open Access
Pedestrian Accessible Infrastructure Inventory: Assessing Zero-Shot Segmentation on Multi-Mode Geospatial Data for All Pedestrian Types
In this paper, a Segment Anything Model (SAM)-based pedestrian infrastructure segmentation workflow is designed and optimized, which is capable of efficiently processing multi-sourced geospatial data including LiDAR data and satellite imagery data. We used an expanded definition of pedestrian infrastructure inventory which goes beyond the traditional transportation elements to include street furniture objects that are important for accessibility but are often omitted from the traditional definition. Our contributions lie in producing the necessary knowledge to answer the following two questions. First, which data representation can facilitate zero-shot segmentation of infrastructure objects with SAM? Second, how well does the SAM-based method perform on segmenting pedestrian infrastructure objects? Our findings indicate that street view images generated from mobile LiDAR point cloud data, when paired along with satellite imagery data, can work efficiently with SAM to create a scalable pedestrian infrastructure inventory approach with immediate benefits to GIS professionals, city managers, transportation owners, and walkers, especially those with travel-limiting disabilities, such as individuals who are blind, have low vision, or experience mobility disabilities.
Open AccessPedestrian Accessible Infrastructure Inventory: Assessing Zero-Shot Segmentation on Multi-Mode Geospatial Data for All Pedestrian Types
In this paper, a Segment Anything Model (SAM)-based pedestrian infrastructure segmentation workflow is designed and optimized, which is capable of efficiently processing multi-sourced geospatial data including LiDAR data and satellite imagery data. We used an expanded definition of pedestrian infrastructure inventory which goes beyond the traditional transportation elements to include street furniture objects that are important for accessibility but are often omitted from the traditional definition. Our contributions lie in producing the necessary knowledge to answer the following two questions. First, which data representation can facilitate zero-shot segmentation of infrastructure objects with SAM? Second, how well does the SAM-based method perform on segmenting pedestrian infrastructure objects? Our findings indicate that street view images generated from mobile LiDAR point cloud data, when paired along with satellite imagery data, can work efficiently with SAM to create a scalable pedestrian infrastructure inventory approach with immediate benefits to GIS professionals, city managers, transportation owners, and walkers, especially those with travel-limiting disabilities, such as individuals who are blind, have low vision, or experience mobility disabilities.
- Open Access
- 5 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 45-49
In-Place Virtual Exploration Using a Virtual Cane: An Initial Study
In this initial study, we addressed the challenge of assisting individuals who are blind or have low vision (BLV) in familiarizing themselves with new environments. Navigating unfamiliar areas presents numerous challenges for BLV individuals. We sought to explore the potential of Virtual Reality (VR) technology to replicate real-world settings, thereby allowing users to virtually experience these spaces at their convenience, often from the comfort of their homes. As part of our preliminary investigation, we designed an interface tailored to facilitate movement for BLV users without needing physical mobility. Our study involved six blind participants. Early findings revealed that participants encountered difficulties adapting to the interface. Post-experiment interviews illuminated the reasons for these challenges, including issues with interface usage, the complexity of managing multiple interface elements, and the disparity between physical movement and interface use. Given the early stage of this research, these findings provide valuable insights for refining the approach and improving the interface in future iterations.
- 5 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 45-49
Open AccessIn-Place Virtual Exploration Using a Virtual Cane: An Initial Study
In this initial study, we addressed the challenge of assisting individuals who are blind or have low vision (BLV) in familiarizing themselves with new environments. Navigating unfamiliar areas presents numerous challenges for BLV individuals. We sought to explore the potential of Virtual Reality (VR) technology to replicate real-world settings, thereby allowing users to virtually experience these spaces at their convenience, often from the comfort of their homes. As part of our preliminary investigation, we designed an interface tailored to facilitate movement for BLV users without needing physical mobility. Our study involved six blind participants. Early findings revealed that participants encountered difficulties adapting to the interface. Post-experiment interviews illuminated the reasons for these challenges, including issues with interface usage, the complexity of managing multiple interface elements, and the disparity between physical movement and interface use. Given the early stage of this research, these findings provide valuable insights for refining the approach and improving the interface in future iterations.