Indu Bodala

Abiram Panchalingam, Indu Bodala, Stuart Middleton

High-fidelity gaze redirection is critical for generating augmented data to improve the generalization of gaze estimators. 3D Gaussian Splatting (3DGS) models like GazeGaussian represent the state-of-the-art but can struggle with rendering subtle, continuous gaze shifts. In this paper, we propose DiT-Gaze, a framework that enhances 3D gaze redirection models using a novel combination of Diffusion Transformer (DiT), weak supervision across gaze angles, and an orthogonality constraint loss. DiT allows higher-fidelity image synthesis, while our weak supervision strategy using synthetically generated intermediate gaze angles provides a smooth manifold of gaze directions during training. The orthogonality constraint loss mathematically enforces the disentanglement of internal representations for gaze, head pose, and expression. Comprehensive experiments show that DiT-Gaze sets a new state-of-the-art in both perceptual quality and redirection accuracy, reducing the state-of-the-art gaze error by 4.1% to 6.353 degrees, providing a superior method for creating synthetic training data. Our code and models will be made available for the research community to benchmark against.

Oliver Grainge, Michael Milford, Indu Bodala et al.

Visual Place Recognition (VPR) is fundamental for the global re-localization of robots and devices, enabling them to recognize previously visited locations based on visual inputs. This capability is crucial for maintaining accurate mapping and localization over large areas. Given that VPR methods need to operate in real-time on embedded systems, it is critical to optimize these systems for minimal resource consumption. While the most efficient VPR approaches employ standard convolutional backbones with fixed descriptor dimensions, these often lead to redundancy in the embedding space as well as in the network architecture. Our work introduces a novel structured pruning method, to not only streamline common VPR architectures but also to strategically remove redundancies within the feature embedding space. This dual focus significantly enhances the efficiency of the system, reducing both map and model memory requirements and decreasing feature extraction and retrieval latencies. Our approach has reduced memory usage and latency by 21% and 16%, respectively, across models, while minimally impacting recall@1 accuracy by less than 1%. This significant improvement enhances real-time applications on edge devices with negligible accuracy loss.

Oliver Grainge, Michael Milford, Indu Bodala et al.

Visual Place Recognition (VPR) is a critical task for performing global re-localization in visual perception systems. It requires the ability to accurately recognize a previously visited location under variations such as illumination, occlusion, appearance and viewpoint. In the case of robotic systems and augmented reality, the target devices for deployment are battery powered edge devices. Therefore whilst the accuracy of VPR methods is important so too is memory consumption and latency. Recently new works have focused on the recall@1 metric as a performance measure with limited focus on resource utilization. This has resulted in methods that use deep learning models too large to deploy on low powered edge devices. We hypothesize that these large models are highly over-parameterized and can be optimized to satisfy the constraints of a low powered embedded system whilst maintaining high recall performance. Our work studies the impact of compact convolutional network architecture design in combination with full-precision and mixed-precision post-training quantization on VPR performance. Importantly we not only measure performance via the recall@1 score but also measure memory consumption and latency. We characterize the design implications on memory, latency and recall scores and provide a number of design recommendations for VPR systems under these resource limitations.

Oliver Grainge, Michael Milford, Indu Bodala et al.

TAT-VPR is a ternary-quantized transformer that brings dynamic accuracy-efficiency trade-offs to visual SLAM loop-closure. By fusing ternary weights with a learned activation-sparsity gate, the model can control computation by up to 40% at run-time without degrading performance (Recall@1). The proposed two-stage distillation pipeline preserves descriptor quality, letting it run on micro-UAV and embedded SLAM stacks while matching state-of-the-art localization accuracy.

Oliver Grainge, Michael Milford, Indu Bodala et al.

Visual Place Recognition (VPR) localizes a query image by matching it against a database of geo-tagged reference images, making it essential for navigation and mapping in robotics. Although Vision Transformer (ViT) solutions deliver high accuracy, their large models often exceed the memory and compute budgets of resource-constrained platforms such as drones and mobile robots. To address this issue, we propose TeTRA, a ternary transformer approach that progressively quantizes the ViT backbone to 2-bit precision and binarizes its final embedding layer, offering substantial reductions in model size and latency. A carefully designed progressive distillation strategy preserves the representational power of a full-precision teacher, allowing TeTRA to retain or even surpass the accuracy of uncompressed convolutional counterparts, despite using fewer resources. Experiments on standard VPR benchmarks demonstrate that TeTRA reduces memory consumption by up to 69% compared to efficient baselines, while lowering inference latency by 35%, with either no loss or a slight improvement in recall@1. These gains enable high-accuracy VPR on power-constrained, memory-limited robotic platforms, making TeTRA an appealing solution for real-world deployment.

Indu P. Bodala, Hatice Gunes

Longitudinal interaction studies with Socially Assistive Robots are crucial to ensure that the robot is relevant for long-term use and its perceptions are not prone to the novelty effect. In this paper, we present a dynamic Bayesian network (DBN) to capture the longitudinal interactions participants had with a teleoperated robot coach (RC) delivering mindfulness sessions. The DBN model is used to study complex, temporal interactions between the participants self-reported personality traits, weekly baseline wellbeing scores, session ratings, and facial AUs elicited during the sessions in a 5-week longitudinal study. DBN modelling involves learning a graphical representation that facilitates intuitive understanding of how multiple components contribute to the longitudinal changes in session ratings corresponding to the perceptions of the RC, and participants relaxation and calm levels. The learnt model captures the following within and between sessions aspects of the longitudinal interaction study: influence of the 5 personality dimensions on the facial AU states and the session ratings, influence of facial AU states on the session ratings, and the influences within the items of the session ratings. The DBN structure is learnt using first 3 time points and the obtained model is used to predict the session ratings of the last 2 time points of the 5-week longitudinal data. The predictions are quantified using subject-wise RMSE and R2 scores. We also demonstrate two applications of the model, namely, imputation of missing values in the dataset and estimation of longitudinal session ratings of a new participant with a given personality profile. The obtained DBN model thus facilitates learning of conditional dependency structure between variables in the longitudinal data and offers inferences and conceptual understanding which are not possible through other regression methodologies.

Indu P. Bodala, Nikhil Churamani, Hatice Gunes

Social robots are starting to become incorporated into daily lives by assisting in the promotion of physical and mental wellbeing. This paper investigates the use of social robots for delivering mindfulness sessions. We created a teleoperated robotic platform that enables an experienced human coach to conduct the sessions in a virtual manner by replicating upper body and head pose in real time. The coach is also able to view the world from the robot's perspective and make a conversation with participants by talking and listening through the robot. We studied how participants interacted with a teleoperated robot mindfulness coach over a period of 5 weeks and compared with the interactions another group of participants had with a human coach. The mindfulness sessions delivered by both types of coaching invoked positive responses from the participants for all the sessions. We found that the participants rated the interactions with human coach consistently high in all aspects. However, there is a longitudinal change in the ratings for the interaction with the teleoperated robot for the aspects of motion and conversation. We also found that the participants' personality traits -- conscientiousness and neuroticism influenced the perceptions of the robot coach.

Yaqi Xie, Indu P Bodala, Desmond C. Ong et al.

In this paper, we present results from a human-subject study designed to explore two facets of human mental models of robots---inferred capability and intention---and their relationship to overall trust and eventual decisions. In particular, we examine delegation situations characterized by uncertainty, and explore how inferred capability and intention are applied across different tasks. We develop an online survey where human participants decide whether to delegate control to a simulated UAV agent. Our study shows that human estimations of robot capability and intent correlate strongly with overall self-reported trust. However, overall trust is not independently sufficient to determine whether a human will decide to trust (delegate) a given task to a robot. Instead, our study reveals that estimations of robot intention, capability, and overall trust are integrated when deciding to delegate. From a broader perspective, these results suggest that calibrating overall trust alone is insufficient; to make correct decisions, humans need (and use) multi-faceted mental models when collaborating with robots across multiple contexts.