Image reconstruction—the process of recovering clear images from incomplete or noisy data—has been advancing rapidly through deep learning. Yet most existing approaches rely on costly supervised training and lack theoretical transparency. A new survey maps the rise of unsupervised deep learning for image reconstruction, from traditional denoising-based priors to modern diffusion models. These methods learn structured visual information directly from unlabeled data, and have achieved impressive performance across various fields, including biomedical imaging and remote sensing. The study shows how unsupervised learning based image reconstruction unites neural network efficiency with solid mathematical foundations to achieve both interpretability and flexibility, offering a blueprint for next-generation imaging systems.

Researchers at the University of Melbourne have developed a new AI-based traffic signal control system called M²SAC that improves both fairness and efficiency at urban intersections. Unlike traditional systems focused only on cars, M²SAC accounts for pedestrians, buses, and other users. A key innovation is the phase mask mechanism, which dynamically adjusts green light timings to reduce delays. Tested on real Melbourne traffic data, the model outperformed existing methods, cutting congestion and balancing traffic flow more equitably. The approach supports smarter, fairer, and more inclusive transport systems for modern cities.

To address this challenge, researchers at Korea Advanced Institute of Science and Technology (KAIST) and Donghai Laboratory developed a new model called ProChunkFormer, which reconstructs vehicle trajectories from sparse and noisy GPS data, enabling more accurate mobility analysis and intelligent transportation planning.

The heterogeneity causes spatiotemporal inconsistencies in multimodal data, posing challenges for existing methods in multimodal feature extraction and alignment. First, in the temporal dimension, the microsecond-level temporal resolution of event data is significantly higher than the millisecond-level resolution of RGB data, resulting in temporal misalignment and making direct multimodal fusion infeasible. To address this issue, the researchers design an Event Correction Module (ECM) that temporally aligns asynchronous event streams with their corresponding image frames through optical-flow-based warping. The ECM is jointly optimized with the downstream object detection network to learn task-ware event representations.

A paper released in the Journal of Bioresources and Bioproducts describes a fully bio-based barrier film that outperforms PET and PE in oxygen blocking while carrying an integrated curcumin indicator. Dialdehyde nanocellulose (DCNF) generated by 1 h periodate oxidation is blended 75% with pristine cellulose nanofibres; vacuum filtration builds a self-cross-linked sheet that is then dip-coated with an ethyl-cellulose/curcumin solution. The resulting 40 g m-2 membrane transmits 79% visible light yet transmits only 6.9 cm3 m-2 d-1 O2 (120-fold improvement over unmodified CNF) and 43 g m-2 d-1 water vapor—comparable to commercial polypropylene. Over 500 min the composite scavenges 91% DPPH radicals and suppresses > 99.9% growth of E. coli and S. aureus. When attached to refrigerated shrimp, the yellow label turns orange-brown as volatilised amines raise local pH, giving an irreversible but easy-to-see spoilage alert within 2 d. The approach requires no metals or synthetic antioxidants, making it a candidate for home-compostable, intelligent food packaging.

A study published online in the Journal of Bioresources and Bioproducts describes a single-step benzoylation protocol that converts UV-prone kenaf into a self-photobleaching, radical-quenching biobased reinforcement. After proving esterification and partial delignification with FT-IR, NMR and TGA, the team exposed modified and raw fibers to 300–400 nm light for 500 h. While untreated kenaf yellowed and lost 96 % of its tensile strength, the stabilized benzoylated variant first whitened as surface micro-cracks scattered light, then retained 65 % of original strength with no further color shift. Density-functional calculations show benzoyl substitution raises the hydrogen-abstraction barrier of lignin phenolics by ~20 kcal mol⁻¹, suppressing the semiquinone radicals that normally propagate oxidation. The approach needs only commodity reagents and existing pulp equipment, offering automotive, textile and packaging industries a scalable route to durable, naturally derived composites.

Infants born with Robin sequence often experience breathing and feeding difficulties caused by an underdeveloped lower jaw.

Developing effective, versatile, and high-precision sensing interfaces remains a crucial challenge in human–machine–environment interaction applications. Despite progress in interaction-oriented sensing skins, limitations remain in unit-level reconfiguration, multiaxial force and motion sensing, and robust operation across dynamically changing or irregular surfaces. Herein, we develop a reconfigurable omnidirectional triboelectric whisker sensor array (RO-TWSA) comprising multiple sensing units that integrate a triboelectric whisker structure (TWS) with an untethered hydro-sealing vacuum sucker (UHSVS), enabling reversibly portable deployment and omnidirectional perception across diverse surfaces. Using a simple dual-triangular electrode layout paired with MXene/silicone nanocomposite dielectric layer, the sensor unit achieves precise omnidirectional force and motion sensing with a detection threshold as low as 0.024 N and an angular resolution of 5°, while the UHSVS provides reliable and reversible multi-surface anchoring for the sensor units by involving a newly designed hydrogel combining high mechanical robustness and superior water absorption. Extensive experiments demonstrate the effectiveness of RO-TWSA across various interactive scenarios, including teleoperation, tactile diagnostics, and robotic autonomous exploration. Overall, RO-TWSA presents a versatile and high-resolution tactile interface, offering new avenues for intelligent perception and interaction in complex real-world environments.

To enhance the electrochemical performance of lithium-ion battery anodes with higher silicon content, it is essential to engineer their microstructure for better lithium-ion transport and mitigated volume change as well. Herein, we suggest an effective approach to control the micropore structure of silicon oxide (SiO_x)/artificial graphite (AG) composite electrodes using a perforated current collector. The electrode features a unique pore structure, where alternating high-porosity domains and low-porosity domains markedly reduce overall electrode resistance, leading to a 20% improvement in rate capability at a 5C-rate discharge condition. Using microstructure-resolved modeling and simulations, we demonstrate that the patterned micropore structure enhances lithium-ion transport, mitigating the electrolyte concentration gradient of lithium-ion. Additionally, perforating current collector with a chemical etching process increases the number of hydrogen bonding sites and enlarges the interface with the SiO_x/AG composite electrode, significantly improving adhesion strength. This, in turn, suppresses mechanical degradation and leads to a 50% higher capacity retention. Thus, regularly arranged micropore structure enabled by the perforated current collector successfully improves both rate capability and cycle life in SiO_x/AG composite electrodes, providing valuable insights into electrode engineering.