Many Multi-Object Tracking (MOT) approaches exploit motion information to associate all the detected objects across frames. However, traditional tracking-by-detection (TBD) methods, relying on the Kalman Filter, often work well in linear motion scenarios but struggle to accurately predict the locations of objects undergoing complex and non-linear movements. To overcome these limitations, we propose ETTrack, a novel motion prediction method with an enhanced temporal motion predictor. Specifically, the motion predictor integrates a transformer model and a Temporal Convolutional Network (TCN) to capture both long-term and short-term motion patterns, and it predicts the future motion of individual objects based on the historical motion information. Additionally, we propose a novel Momentum Correction Loss function that provides additional information regarding the motion direction of objects during training. This allows the motion predictor to rapidly adapt to sudden motion variations and more accurately predict future motion. Our experimental results demonstrate that ETTrack achieves a competitive performance compared with state-of-the-art trackers on DanceTrack and SportsMOT, scoring 56.4 $$\%$$% and 74.4 $$\%$$% in HOTA metrics, respectively. Our work provides a robust solution for MOT in complex dynamic environments, which enhances the non-linear motion prediction capabilities of tracking algorithms.

Deep learning (DL)-based Human Activity Recognition (HAR) using wearable inertial measurement unit (IMU) sensors can revolutionize continuous health monitoring and early disease prediction. However, most DL HAR models are untested in their robustness to real-world variability, as they are trained on limited lab-controlled data. In this study, we isolated and analyzed the effects of the subject, device, position, and orientation variabilities on DL HAR models using the HARVAR and REALDISP datasets. The Maximum Mean Discrepancy (MMD) was used to quantify shifts in the data distribution caused by these variabilities, and the relationship between the distribution shifts and model performance was drawn. Our HARVAR results show that different types of variability significantly degraded the DL model performance, with an inverse relationship between the data distribution shifts and performance. The compounding effect of multiple variabilities studied using REALDISP further underscores the challenges of generalizing DL HAR models to real-world conditions. Analyzing these impacts highlights the need for more robust models that generalize effectively to real-world settings. The MMD proved valuable for explaining the performance drops, emphasizing its utility in evaluating distribution shifts in HAR data.

Deep Learning (DL) has enabled considerable increases in the accuracy of classification tasks in several domains, including Human Activity Recognition (HAR). It is well-known that when data distribution changes between the training and test datasets, the accuracy can drop, sometimes significantly. However, some variability sources in HAR, such as sensor orientation, are only sometimes considered when evaluating these models. Therefore, we must understand how much such changes could impact current DL architectures. In this paper, under an orientation variability scenario, we evaluate three common DL architectures, DeepConvLSTM, TinyHAR, and Attend-and-Discriminate, to quantify the performance drop attributed to this shift. Our results show that all architectures show performance drops on average, as expected, but participants are affected differently from them, so they would fall short for some in classification accuracy in real-life settings where orientation can change across the wearing sessions of one participant or across participants. The performance change is related to the difference in distribution distance.

The scarcity of high-quality labeled data in sensor-based Human Activity Recognition (HAR) hinders model performance and limits generalization across real-world scenarios. Data augmentation is a key strategy to mitigate this issue by enhancing the diversity of training datasets. Signal Transformation-based Data Augmentation (STDA) techniques have been widely used in HAR. However, these methods are often physically implausible and can produce data misaligned with activity labels. In this study, we propose Physically Plausible Data Augmentation (PPDA) enabled by physics simulation. PPDA leverages human body movement data from motion capture or video-based pose estimation and incorporates realistic variability through physics simulation, including changes in body movements, sensor placements, and hardware-related effects. We compare the performance of PPDA methods with traditional STDA methods on three public datasets of activities of daily living and fitness workouts: REALDISP (34 classes), REALWORLD (8 classes), and MM-Fit (11 classes). First, we compare each PPDA method with its closest STDA counterpart, showing that PPDA improves macro F1 scores by an average of 3.7 percentage points (pp), with gains up to 13 pp. Second, we assess to what extent combining PPDAs can reduce the need for initial data collection, and in comparison with combined STDAs, we observe that PPDA achieves competitive performance with up to 60% fewer training subjects. As the first systematic study of PPDA in HAR, these results highlight the advantages of pursuing physical plausibility in data augmentation and the potential of physics simulation for generating synthetic Inertial Measurement Unit (IMU) data for training deep learning HAR models. This cost-effective and scalable approach therefore helps address the annotation scarcity challenge in HAR.

Crop height is a vital indicator of growth conditions. Traditional drone image-based crop height measurement methods primarily rely on calculating the difference between the Digital Elevation Model (DEM) and the Digital Terrain Model (DTM). The calculation often needs more ground information, which remains labour-intensive and time-consuming. Moreover, the variations of terrains can further compromise the reliability of these ground models. In response to these challenges, we introduce G-DMD, a novel method based on Gated Recurrent Units (GRUs) using DEM and multispectral drone images to calculate the crop height. Our method enables the model to recognize the relation between crop height, elevation, and growth stages, eliminating reliance on DTM and thereby mitigating the effects of varied terrains. We also introduce a data preparation process to handle the unique DEM and multispectral image. Upon evaluation using a cotton dataset, our G-DMD method demonstrates a notable increase in accuracy for both maximum and average cotton height measurements, achieving a 34% and 72% reduction in Root Mean Square Error (RMSE) when compared with the traditional method. Compared to other combinations of model inputs, using DEM and multispectral drone images together as inputs results in the lowest error for estimating maximum cotton height. This approach demonstrates the potential of integrating deep learning techniques with drone-based remote sensing to achieve a more accurate, labour-efficient, and streamlined crop height assessment across varied terrains.