Conventional RGB stereo matching suffers from ambiguities caused by metamerism and weak textures. We demonstrate that implicitly recovered spectral information can enhance depth estimation without requiring spectral sensors at deployment. A reconstruction-to-matching architecture maps RGB inputs to a latent spectral space, injecting material priors into stereo matching. To address data scarcity, we develop a co-aperture system combining a liquid crystal tunable filter with Scheimpflug LiDAR for pixel-aligned multimodal acquisition, and propose a “Measure-and-Complete” strategy using sparse LiDAR to generate dense pseudo-ground truth. Experiments show 4.34% reduction in endpoint error, validating the effectiveness of spectral priors for geometric perception.

Malaria-attributable mortality remains heavily concentrated among children under five years of age, highlighting a persistent and profound vulnerability within this demographic group. Albeit RTS,S/AS01 provides only modest efficacy, the development of safe, high-efficacy, and cost-effective vaccines becomes imperative. Consequently, novel prevention strategies—particularly antibody-based interventions—emerges as a critical priority for reducing global malaria morbidity and mortality. Here we use an active learning–driven XGBoost framework to enable predictive antibody profiles discovery of antibody features attributed to malaria protection. Our active learning technique integrates detailed physicochemical feature engineering from heavy and light chain CDR3 sequences with participant-level cross-validation to prevent data leakage from within-donor correlations. By iteratively selecting the most informative sequences for labeling, our computational framework significantly reduces experimental burden (44.4%) and is 1.8x more efficient, while maintaining high predictive performance. The resulting model operates on an ultra-efficient computational architecture, yielding protection-state predictions with optimal performance across diverse antibody repertoires. To assess scalability, we extend the active-learning pipeline to larger IGHV–IGHJ feature sets and higher-dimensional repertoire landscapes, achieving robust generalization across trial regimen. Our findings enhance the classification performance by discriminating protection status from underlying antibody profiles while jointly modeling time- and dose-dependent effects. Our active learning-based inference reaches AUPRC of  ∼ 0.9385 with 8503 labeled sequences, compared with 15,305 sequences required under random sampling, indicating a substantial improvement in label efficiency. This approach demonstrates that active-learning–guided gradient-boosted inference can support large-scale immune profiling with minimal labeled data, paving the way for rapid, cost-effective ( > 44%) computational screening strategies for next-generation malaria vaccines and antibody-based interventions.

Infrared optical wireless communication (OWC) is a promising candidate for next-generation wireless links due to its high bandwidth and inherent security. However, the limited field-of-view (FOV) of terminals induces great difficulty in establishing line-of-sight (LoS) link between the transmitter and the terminal. In this paper, to facilitate efficient beam alignment, we report an auxiliary strategy integrating an external angle-sensing module with a wide-FOV telecentric receiver architecture, providing the receiver with real-time incident of angle (IoA) information to transform the traditional blind search into a deterministic alignment process. Three different angle-sensing modalities-monocular vision, binocular stereo vision, and LiDAR scanning-are systematically evaluated. Experimental results indicate that the precision of vision-based methods degrades significantly when the detection distance exceeds 1.3 m, whereas the LiDAR-based scheme maintains an angular error below 0.714 degrees across a 3 m range and a 100 degrees FOV, demonstrating superior robustness and reliability. By leveraging the LiDAR-derived IoA, the receiver implements a "coarse-to-fine" two-stage alignment mechanism. Combined with a non-coaxial correction algorithm, this approach bypasses the need for redundant internal search modules and bulky fiber bundles, enabling a single fiber to achieve micron-level auto-coupling with a peak steady-state efficiency of 54.2%. Under these conditions, a 10 Gbps on-off-keying (OOK)-modulated OWC link is successfully maintained within a 100 degrees FOV, with the bit error rate (BER) consistently meeting the forward error correction (FEC) requirement. This research demonstrates that sensor-aided alignment simplifies terminal architecture and accelerates link acquisition, providing a viable pathway for wide-FOV optical wireless networks.

Accurate assessment of knee osteoarthritis (KOA) is limited by the current arthroscopy's reliance on subjective visual inspection. We propose a 3D spatio-spectral automatic grading framework combining deep learning with chromatic differential confocal matrix-based 3D topography and hyperspectral imaging (CDCM-THI). Our digital micromirror device-based system simultaneously captures high-resolution 3D micro-topography and hyperspectral cartilage signatures (lateral: 3.91 & micro;m, axial: 1.96 & micro;m, spectral: similar to 15 nm). To analyze this multi-modal data, we developed an improved ResNet50 late fusion network optimized with a consistency-enhanced loss function. Evaluated on ex vivo clinical samples, our framework achieved an average cross-validation accuracy of 94.83%, significantly outperforming single-modal baselines. This non-destructive approach bridges the gap between topographical assessment and broadband optical analysis, providing a powerful quantitative tool for future deep learning-assisted arthroscopic integration.

The accumulation of various colored plastics waste in terrestrial ecosystems poses an escalating threat to environmental sustainability, demanding effective and precise classification methods. However, current machine vision techniques often fail due to significant interference caused by plastic colorants. Their performance in real-world scenarios is further compromised by complex soil backgrounds, including soil heterogeneity and varying moisture content. To overcome these limitations, we present a novel NIR hyperspectral imaging system, integrated with a specially designed MS3D-Net, for high-precision classification of soil-embedded plastics. The MS3D-Net model simultaneously performs pixel-level segmentation and polymer classification for the identification of 11 plastic types within complex soil backgrounds. We evaluated the system's robustness across three distinct heterogeneous soil types (loam, clay, and sand) under varying moisture gradients, demonstrating stable performance against matrix interference. To validate applicability in realistic field conditions, the system was tested on plastics subjected to aging, oxidation, fragmentation, and biofouling embedded in natural soils containing complex impurities such as plant roots and residues, achieving an average precision of 0.89. The method successfully recognized irregular consumer plastic products collected from the environment. Our approach provides a robust solution for identifying diverse plastic polymers in complex soil environments, offering technical support for environmental sustainability and plastic waste management.

A training-free, complex-valued differentiable focusing framework for digital holography is established utilizing band-limited angular spectrum propagation and Wirtinger gradient flow. By optimizing propagation distance as a continuous variable and introducing spatial-domain metrics (e.g., total variation), this method explicitly overcomes the zero-gradient limitation of frequency-domain quadratic criteria to ensure stable analytical gradients. A coarse-to-fine strategy then generates z-encoded multifocal complex fields for coherent extended-depth-of-field reconstruction. Quantitative validation shows that DHF-Frame-RLG localizes the optimal reconstruction plane with an axial MAE of 1.0 µm and RMSE of 1.8 µm, and achieves a 96% autofocus success rate, corresponding to a 4.5× reduction in evaluations relative to exhaustive search. Crucially, this framework mitigates boundary blurring and wavefront coherence loss caused by traditional discrete axial scans, providing a robust physical foundation for future integration with physics-informed learning models.

An ultra-enhanced near-infrared (NIR) photoacoustic gas sensor was developed by integrating three enhancing techniques: (a) boosting the excitation power up to 2 W via a custom-built large-mode erbium doped fiber amplifier (EDFA), (b) exploiting the acoustic resonance amplification of a hyperbolic nonlinear resonator (HNR), and (c) increasing the effective absorption path length by using a near-concentric multipass cavity (MPC) with 20 reflections. A weak CO absorption line at 1566.64 nm with the intensity of 2.074 × 10^-23 cm/molecule was selected. The photoacoustic signal was enhanced 396 times. A minimum detection limit (MDL) of 190 ppb at 10 s was achieved and can be improved to be 11.4 ppb according to the Allan analysis, which was comparable to a mid-infrared (MIR) photoacoustic sensor. The ultra-enhanced NIR photoacoustic sensor is a cost-effective solution for the ppb-level trace gas detection, offering a price that is less than one-third that of MIR photoacoustic sensors. 

High-quality GaN epitaxy and transfer are essential for flexible optoelectronics. Yet, the trade-off between strong covalent-bonded interfaces for epitaxy and weak forces for exfoliation hinders the simultaneous achievement of high-quality epitaxy and intact transfer. To address this, we report a dual-functional metal-interlayer chemical lift-off (MI-CLO) strategy. To the best of our knowledge, this is the first demonstration of a metallic interlayer serving as both an epitaxial template and a dissolvable sacrificial layer, facilitating both single-crystalline GaN thin film growth and low-damage transfer to flexible substrates. Our results demonstrate that the copper (Cu) interlayer yields superior surface morphology and crystalline integrity of the epilayer than nickel, despite their structural similarities. By incorporating a Cu interlayer and composite buffer layers, we obtain a 33-µm GaN film on sapphire with the (0002) rocking curve full width half maximum as low as 0.10°. A 5-µm GaN thin film exhibits a low threading dislocation density (8 × 10⁸ cm⁻²) comparable to direct heteroepitaxial GaN films on sapphire. The MI-CLO technology demonstrates broad applicability across representative heteroepitaxial systems, facilitating strain-released transfer with high epilayer crystalline integrity onto flexible substrates. A flexible ultraviolet photodetector is demonstrated, maintaining stable responses under mechanical bending and verifying its potential for flexible optoelectronics.

Applications of bound states in the continuum (BIC) are limited by the challenges of dynamical tuning and coupling to external excitation. In this work, a well-thought-out structure model of two gratings coupled by a waveguide is reported. The structure results in unique BIC-Fano dual modes which undergo an unusual degeneration and inversion by either the tuning of structure parameters or through misalignment of the bilayer gratings, accompanied by the merging and splitting of Quasi(Q)-BIC and Fano peaks in the transport spectra of the structure. The BIC-Fano dual mode degeneration leads to the infinite Q factor from BIC and the coupling channel to external excitation from resonance. The results suggest an elegant way for smart engineering of resonances for many applications in optical switching and enhanced light−matter interactions. Especially, the dual-mode structure gives a highly efficient coupling to a planar waveguide, and is also useful for displacement sensing with an extremely high resolution.

Designing efficient electrocatalysts for hydrogen generation remains a major challenge because the hydrogen evolution reaction (HER) still relies heavily on platinum (Pt). At the same time, most non-Pt alternatives are limited to single-metal or symmetric bimetallic active sites. Conductive metal–organic frameworks (MOFs) with conjugated ligands are attractive candidates because their redox-active metal nodes can enhance electrocatalytic performance. In this work, we designed and synthesized three H₃TATB-based (2,4,6-tris(4-carboxyphenyl)-s-triazine) MOFs, Cu-MOF, Ni-MOF, and asymmetric bimetallic CuNi-MOF, for HER in alkaline media. Among them, CuNi-MOF exhibited the best performance, requiring only 123 mV overpotential to achieve 10 mA cm⁻² with a Tafel slope of 150 mV dec⁻¹, outperforming Ni-MOF and Cu-MOF. Its enhanced activity is attributed to cooperative interactions between Cu and Ni, which optimize the electronic structure, enhance charge transfer, and accelerate HER kinetics while maintaining stability for 24 h. The triazine-based ligand further stabilizes the active sites and promotes electron transport and hydrogen intermediate adsorption. DFT calculations reveal a narrow bandgap of 0.684 eV and a favorable heteronuclear CuNi configuration that lowers the HER energy barrier. This strategy provides a flexible platform for developing multi-metallic π-conjugated MOFs for alkaline HER.