Zhaoyang Gu

Exercise recognition and movement quality assessment remain challenging in supervised exercise training, particularly under viewpoint changes and self-occlusion. Vision-based methods provide spatial posture information but are susceptible to keypoint loss, whereas inertial sensing is less affected by occlusion but provides limited information about global posture geometry. This study presents a dual-stream prototype that combines a nine-axis inertial measurement unit (IMU) with vision-based pose estimation. A 1DCNN-LSTM branch models inertial dynamics, a custom keypoint temporal branch models normalized pose sequences, and a confidence-gated rule adjusts their contributions according to visual keypoint reliability. Owing to the absence of a public synchronized multi-view IMU–vision exercise dataset with the required protocol, we constructed IMV-Exercise, comprising 10 participants, three exercises, and 900 repetition-level samples with side-, front-, and posterior-view recordings. The system achieved 96.0% exercise recognition accuracy under leave-one-subject-out cross-validation. In a separate viewpoint-specific evaluation, the fused output achieved 91.2% action-window accuracy under posterior viewing. Across 50 online trials, the reported recognition accuracy was 96.0%, the mean end-to-end latency was 195 ms, and the recorded maximum was below 210 ms. These results establish feasibility within the studied cohort and exercises; broader generalization and feedback effectiveness require larger, independently controlled evaluations.

Information analysis recommendation differs from conversational recommender systems (CRS) because relevance changes with the decision phase. The same event may support observation, interpretation, option selection, or action feedback, yet most large language model (LLM)-agent CRS represent dialogue state as intent and preference. This study examines whether explicit decision-stage awareness improves recommendation and whether it can be added independently of the LLM backbone. We propose Stage-Aware Conversational Recommender System (SA-CRS), a plug-in layer guided by the Observe–Orient–Decide–Act cycle. It decouples stage detection from LLM reasoning and uses detected stages to guide dialogue strategy and candidate re-ranking. We evaluate SA-CRS on an information analysis recommendation dataset from event-structured reports, using multi-turn simulated dialogues and four LLM backbones. Oracle stage injection improves Hit@5 by 3.0 percentage points (pp), showing that decision stage provides a signal beyond topic matching. With a prompt-based detector, SA-CRS improves Hit@5 by 9.0 pp on a strong backbone; with an independent Bidirectional Encoder Representations from Transformers (BERT) detector and probabilistic re-ranking, gains range from 6.5 to 15.5 pp. Negative controls with uniform or random stage signals fail to reproduce the improvements and may reduce efficiency. The results suggest that, within the evaluated single-domain information-analysis setting, decoupled decision-stage awareness is practical for decision-intensive CRS.

Accurate extraction of road networks from high-resolution remote sensing imagery is a fundamental task underpinning autonomous-driving navigation, urban spatial planning, and the dynamic updating of geographic information databases. Although existing road extraction methods attain outstanding pixel-level segmentation accuracy and topological integrity, most follow an accuracy-first design paradigm that relies on heavyweight backbones and increasingly complex decoders, incurring a parameter volume and storage overhead that constitute the principal bottleneck for deploying them on resource-constrained edge platforms such as unmanned aerial vehicles, mobile terminals, and onboard satellite processors. Conversely, models that pursue extreme lightweighting often fail to preserve the thin, continuous, linear structure of roads, tending to produce topological breaks in the extracted road networks. To bridge the performance gap between segmentation accuracy and model size, we propose LOA-Net, a lightweight orientation-aware road extraction network. LOA-Net introduces a Road-Aligned Deformable Convolution (RA-DCN) that adaptively aligns the sampling region with the road geometry and explicitly supervises the predicted road orientation, thereby accurately capturing road connectivity while substantially reducing the parameter count. Experiments on the CHN6-CUG and DeepGlobe benchmarks show that LOA-Net surpasses representative state-of-the-art methods on both IoU and F1, while achieving the lowest parameter count of all compared models and a computational complexity comparable to its peers, striking an excellent trade-off between segmentation performance and a mobile-friendly footprint that makes it well suited for road extraction from remote sensing imagery in resource-constrained scenarios.

Building change detection from bi-temporal remote-sensing imagery underpins urban planning, infrastructure monitoring, and disaster assessment. Existing deep-learning methods achieve high accuracy but rely on large parameter counts, while pixel-level supervision provides limited boundary guidance. We propose POCA-lite, a lightweight encoder–decoder with an inference-coupled geometry branch: three geometric prediction heads—distance transform, boundary, and center heatmap—whose outputs are fused back into the decoder via a feedback pathway active at both training and inference. On the LEVIR-CD benchmark under a unified retraining protocol, multi-seed evaluation shows that POCA-lite matches SNUNet in mean F1 while using 47% fewer parameters and 53% fewer FLOPs. Boundary F1 improves by 9.22 pp over the no-geometry baseline. Decomposition ablations reveal two complementary improvement sources: geometric supervision alone recovers 85% of the total gain, while the feedback fusion pathway recovers 92%; their combination achieves the full result. Geometry-aware targets outperform a generic multitask control. Cross-architecture transfer to SNUNet yields +1.06 pp F1. However, cross-dataset evaluation on WHU-CD shows that the method underperforms SNUNet on dense urban morphology, and zero-shot cross-dataset transfer is not established. These results indicate that inference-coupled geometric supervision is effective for lightweight, boundary-sensitive change detection on domains with well-separated building morphology, but its applicability is scope-bounded.

Building change detection from bi-temporal remote sensing imagery underpins urban planning, infrastructure monitoring, and disaster assessment. Existing deep-learning methods achieve high accuracy but rely on large parameter counts, while pixel-level supervision provides limited boundary guidance. We propose POCA-lite, a lightweight encoder-decoder with an inference-coupled geometry branch: three geometric prediction heads—distance transform, boundary, and center heatmap—whose outputs are fused back into the decoder via a feedback pathway active at both training and inference. On the LEVIR-CD benchmark under a unified retraining protocol, multi-seed evaluation shows that POCA-lite matches SNUNet in mean F1 while using 47% fewer parameters and 53% fewer FLOPs. Boundary F1 improves by 9.22 pp over the no-geometry baseline. Decomposition ablations reveal two complementary improvement sources: geometric supervision alone recovers 85% of the total gain, while the feedback fusion pathway recovers 92%; their combination achieves the full result. Geometry-aware targets outperform a generic multi-task control. Cross-architecture transfer to SNUNet yields +1.06 pp F1. However, cross-dataset evaluation on WHU-CD shows that the method underperforms SNUNet on dense urban morphology, and zero-shot cross-dataset transfer is not established. These results indicate that inference-coupled geometric supervision is effective for lightweight, boundary-sensitive change detection on domains with well-separated building morphology, but its applicability is scope-bounded.

Unsupervised change detection (UCD) from heterogeneous bitemporal optical–SAR imagery is challenging due to modality discrepancy, speckle/illumination variations, and the absence of change annotations. We propose MV-S2CD, a vision foundation model (VFM)-based framework that learns a modality-bridged latent space and produces dense change maps in a fully unsupervised manner. To robustly adapt pretrained VFM priors to heterogeneous inputs with minimal task-specific parameters, MV-S2CD incorporates lightweight modality-specific adapters and parameter-efficient low-rank adaptation (LoRA) in high-level layers. A shared projector embeds the two observations into a common geometry, enabling consistent cross-modal comparison and reducing sensor-induced domain shift. Building on the bridged representation, we design a dual-branch change reasoning module that decouples structure-sensitive cues from semantic-consistency cues: a structure pathway preserves fine boundaries and local variations, while a semantic-consistency pathway employs reliability gating and multi-scale context aggregation to suppress pseudo-changes caused by modality-specific nuisances and residual misregistration. For label-free optimization, we develop a difference-centric self-supervision scheme with two perturbation views and reliability-guided pseudo-partitioning, jointly enforcing pseudo-unchanged invariance, pseudo-changed/unchanged separability, and sparsity and edge-preserving regularization. Experiments on three heterogeneous optical–SAR benchmarks demonstrate that MV-S2CD consistently improves the Precision–Recall trade-off and achieves state-of-the-art performance among unsupervised baselines, while remaining backbone-flexible and efficient.

Unsupervised change detection (UCD) from heterogeneous bitemporal optical–SAR imagery is challenging due to modality discrepancy, speckle/illumination variations, and the absence of change annotations. We propose MV-S2CD, a vision foundation model (VFM)-based framework that learns a modality-bridged latent space and produces dense change maps in a fully unsupervised manner. To robustly adapt pretrained VFM priors to heterogeneous inputs with minimal task-specific parameters, MV-S2CD incorporates lightweight modality-specific adapters and parameter-efficient low-rank adaptation (LoRA) in high-level layers. A shared projector embeds the two observations into a common geometry, enabling consistent cross-modal comparison and reducing sensor-induced domain shift. Building on the bridged representation, we design a dual-branch change reasoning module that decouples structure-sensitive cues from semantic-consistency cues: a structure pathway preserves fine boundaries and local variations, while a semantic-consistency pathway employs reliability gating and multi-scale context aggregation to suppress pseudo-changes caused by modality-specific nuisances and residual misregistration. For label-free optimization, we develop a difference-centric self-supervision scheme with two perturbation views and reliability-guided pseudo partitioning, jointly enforcing pseudo-unchanged invariance, pseudo-changed/unchanged separability, and sparsity and edge-preserving regularization. Experiments on three heterogeneous optical–SAR benchmarks demonstrate that MV-S2CD consistently improves the precision–recall trade-off and achieves state-of-the-art performance among unsupervised baselines, while remaining backbone-flexible and efficient.