MOODv2: Masked Image Modeling for Out-of-Distribution Detection.

The crux of effective out-of-distribution (OOD) detection lies in acquiring a robust in-distribution (ID) representation, distinct from OOD samples. While previous methods predominantly leaned on recognition-based techniques for this purpose, they often resulted in shortcut learning, lacking comprehensive representations. In our study, we conducted a comprehensive analysis, exploring distinct pretraining tasks and employing various OOD score functions. The results highlight that the feature representations pre-trained through reconstruction yield a notable enhancement and narrow the performance gap among various score functions. This suggests that even simple score functions can rival complex ones when leveraging reconstruction-based pretext tasks. Reconstruction-based pretext tasks adapt well to various score functions. As such, it holds promising potential for further expansion. Our OOD detection framework, MOODv2, employs the masked image modeling pretext task. Without bells and whistles, MOODv2 impressively enhances 14.30% AUROC to 95.68% on ImageNet and achieves 99.98% on CIFAR-10.

MOODv2: Masked Image Modeling for Out-of-Distribution Detection.

The crux of effective out-of-distribution (OOD) detection lies in acquiring a robust in-distribution (ID) representation, distinct from OOD samples. While previous methods predominantly leaned on recognition-based techniques for this purpose, they often resulted in shortcut learning, lacking comprehensive representations. In our study, we conducted a comprehensive analysis, exploring distinct pretraining tasks and employing various OOD score functions. The results highlight that the feature representations pre-trained through reconstruction yield a notable enhancement and narrow the performance gap among various score functions. This suggests that even simple score functions can rival complex ones when leveraging reconstruction-based pretext tasks. Reconstruction-based pretext tasks adapt well to various score functions. As such, it holds promising potential for further expansion. Our OOD detection framework, MOODv2, employs the masked image modeling pretext task. Without bells and whistles, MOODv2 impressively enhances 14.30% AUROC to 95.68% on ImageNet and achieves 99.98% on CIFAR-10.

BAL: Balancing Diversity and Novelty for Active Learning.

The objective of Active Learning is to strategically label a subset of the dataset to maximize performance within a predetermined labeling budget. In this study, we harness features acquired through self-supervised learning. We introduce a straightforward yet potent metric, Cluster Distance Difference, to identify diverse data. Subsequently, we introduce a novel framework, Balancing Active Learning (BAL), which constructs adaptive sub-pools to balance diverse and uncertain data. Our approach outperforms all established active learning methods on widely recognized benchmarks by 1.20%. Moreover, we assess the efficacy of our proposed framework under extended settings, encompassing both larger and smaller labeling budgets. Experimental results demonstrate that, when labeling 80% of the samples, the performance of the current SOTA method declines by 0.74%, whereas our proposed BAL achieves performance comparable to the full dataset.

Adaptive Perspective Distillation for Semantic Segmentation.

Strong semantic segmentation models require large backbones to achieve promising performance, making it hard to adapt to real applications where effective real-time algorithms are needed. Knowledge distillation tackles this issue by letting the smaller model (student) produce similar pixel-wise predictions to that of a larger model (teacher). However, the classifier, which can be deemed as the perspective by which models perceive the encoded features for yielding observations (i.e., predictions), is shared by all training samples, fitting a universal feature distribution. Since good generalization to the entire distribution may bring the inferior specification to individual samples with a certain capacity, the shared universal perspective often overlooks details existing in each sample, causing degradation of knowledge distillation. In this paper, we propose Adaptive Perspective Distillation (APD) that creates an adaptive local perspective for each individual training sample. It extracts detailed contextual information from each training sample specifically, mining more details from the teacher and thus achieving better knowledge distillation results on the student. APD has no structural constraints to both teacher and student models, thus generalizing well to different semantic segmentation models. Extensive experiments on Cityscapes, ADE20K, and PASCAL-Context manifest the effectiveness of our proposed APD. Besides, APD can yield favorable performance gain to the models in both object detection and instance segmentation without bells and whistles.

Advanced biomaterials in cell preservation: Hypothermic preservation and cryopreservation.

Cell-based medicine has made great advances in clinical diagnosis and therapy for various refractory diseases, inducing a growing demand for cell preservation as support technology. However, the bottleneck problems in cell preservation include low efficiency and poor biocompatibility of traditional protectants. In this review, cell preservation technologies are categorized according to storage conditions: hypothermic preservation at 1 °C~35 °C to maintain short-term cell viability that is useful in cell diagnosis and transport, while cryopreservation at -196 °C~-80 °C to maintain long-term cell viability that provides opportunities for therapeutic cell product storage. Firstly, the background and developmental history of the protectants used in the two preservation technologies are briefly introduced. Secondly, the progress in different cellular protection mechanisms for advanced biomaterials are discussed in two preservation technologies. In hypothermic preservation, the hypothermia-induced and extracellular matrix-loss injuries to cells are comprehensively summarized, as well as the recent biomaterials dependent on regulation of cellular ATP level, stabilization of cellular membrane, balance of antioxidant defense system, and supply of mimetic ECM to prolong cell longevity are provided. In cryopreservation, cellular injuries and advanced biomaterials that can protect cells from osmotic or ice injury, and alleviate oxidative stress to allow cell survival are concluded. Last, an insight into the perspectives and challenges of this technology is provided. We envision advanced biocompatible materials for highly efficient cell preservation as critical in future developments and trends to support cell-based medicine. STATEMENT OF SIGNIFICANCE: Cell preservation technologies present a critical role in cell-based applications, and more efficient biocompatible protectants are highly required. This review categorizes cell preservation technologies into hypothermic preservation and cryopreservation according to their storage conditions, and comprehensively reviews the recently advanced biomaterials related. The background, development, and cellular protective mechanisms of these two preservation technologies are respectively introduced and summarized. Moreover, the differences, connections, individual demands of these two technologies are also provided and discussed.

A Zwitterionic-Aromatic Motif-Based ionic skin for highly biocompatible and Glucose-Responsive sensor.

Electronic skins that can sense external stimuli have been of great significance in artificial intelligence and smart wearable devices in recent years. However, most of current skin materials are unable to achieve high biocompatibility and anti-bacterial activity, which are particularly critical to wearable sensors for neonatal/premature monitoring or tissue-interfaced biosensors (such as electronic wound dressing and smart contact lens). Herein, a zwitterionic-aromatic motif-based conductive hydrogel with electrostatic and π-π interactions is designed for the development of ionic skin sensors. The hydrogel possesses high biocompatibility, anti-bacterial activity, especially glucose-responsive property which has not been achieved by previous ionic skins. Due to its unique molecular design, the zwitterionic-aromatic skin sensor exhibits excellent mechanical properties (robust elasticity and large stretchability) and high-sensitive pressure detection (including a gentle finger touch, small water droplets, and vocal cord vibration). More importantly, aromatic motives in phenylboronic acid segments endow the skin with glucose-responsive property. This skin sensor not only shows great potential in wearable e-skins, but also possesses a promising property for the tissue-interfaced and implantable continuous-glucose-monitor biosensors such as smart wound dressing with a high demand of biocompatibility.

Amphiphilic Marine Antifouling Coatings Based on a Hydrophilic Polyvinylpyrrolidone and Hydrophobic Fluorine-Silicon-Containing Block Copolymer.

The development of environmentally friendly and highly efficient antifouling coatings is vastly desirable in the marine industry. Herein, we prepared a novel amphiphilic block copolymer that combined hydrophilic polyvinylpyrrolidone (PVP) with hydrophobic poly(1-(1H,1H,2H,2H-perfluorodecyloxy)-3-(3,6,9-trioxadecyloxy)-propan-2-yl acrylate) (PFA) and polydimethylsiloxane (PDMS). The amphiphilic copolymer (PVP-PFA-PDMS) was blended into a cross-linked PDMS matrix to form a set of controlled surface composition and surface-renewal coatings with efficient antifouling and fouling-release properties. These coatings incorporated the biofouling settlement resistance ability attributed to the hydrophilic PVP segments and the reduced adhesion strength attributed to the low surface energy of fluorine-silicon-containing segments. As expected, the coatings showed an excellent antifouling performance against bacteria and marine unicellular Navicula parva diatoms (98.1 and 98.5% of reduction, respectively) and fouling-release performance against pseudobarnacle adhesion (84.3% of reduction) compared to the pristine PDMS coating. Moreover, a higher-content PVP-based coatings presented higher ability to resist biofouling adhesion. The nontoxic antifouling coating developed in this paper hold the potential to be applied in a variety of marine industrial facilities.

In Situ Encapsulation of Postcryopreserved Cells Using Alginate Polymer and Zwitterionic Betaine.

Currently, the state-of-the-art cryoprotectants for cell cryopreservation have bottleneck problems (such as cytotoxicity), which place enormous logistical limitations to the development of regenerative medicine. In this work, a first alginate polymer-based approach for human chondrocyte cryopreservation is reported. Combined with zwitterionic betaine, a natural osmoprotectant to offer intracellular protection, this alginate polymer-based approach can achieve ∼90% cryopreservation efficiency. Because of the biocompatibility of alginate polymer and betaine, this approach can easily retrieve the post-thaw cells without traditional multistep cryoprotectant washing procedures, which is highly favorable to cell therapy. Meanwhile, because of the feasible and mild gelation process of alginate polymer, this approach can also directly encapsulate the post-thaw cells into hydrogels without cryoprotectant removal, which is highly useful to tissue engineering. Moreover, these hydrogels exhibit tunable mechanical properties and can form variable shapes and sizes of scaffolds to inject into the patient's defect sites. After encapsulating post-thaw cells, these hydrogels can maintain high cell viability (∼90%) and normal cellular functions for at least 14 days. This work provides a step-change in cryopreservation of cells to be directly used in cell-based applications and may realize promising cellular therapy products that can integrate preservation with clinical practice.