In this paper, we address the novel task of egocentric modality generalization action recognition, which aims to learn a unified discrete representation from paired multimodal egocentric action data during pre-training. This approach enables cross-modal zero-shot generalization in downstream tasks, where the modalities available during inference and training are disjoint. While recent efforts have focused on aligning instance-level or temporal features to reduce feature distribution discrepancies across modalities, they have overlooked the inherent structural categorization within action data. To address this limitation, we propose Modal-Agnostic Prototype Learning (MAPLE), a framework that leverages a prototype memory bank to capture categorical structures. This is further enhanced by a robust semantic disentanglement module and a moment aggregation mechanism, enabling semantically similar behaviors to cluster more closely in the latent space and promoting robust cross-modal generalization. Extensive experiments on the Ego4D and WEAR datasets demonstrate that MAPLE significantly outperforms existing approaches, marking a substantial advancement in the field of egocentric action recognition.

The minimum storage rack-aware regenerating (MSRR) code is a variation of regenerating codes that achieves the optimal repair bandwidth for a single node failure in the rack-aware model. Some explicit constructions of MSRR codes for all parameters to repair a single failed node have been reported. This paper studies MSRR codes with error-correcting capability for multiple erasure tolerance. First, we propose a general repair model of maximum distance separable (MDS) codes with error-correcting capability for multiple erasure tolerance and derive lower bounds on the number of symbols downloaded and accessed, respectively from helper racks for the purpose of correction and repair. Then, we construct a class of MDS array codes and scalar Reed-Solomon (RS) codes with the optimal repair bandwidth and error resilient capability for multiple node failures. Further, our codes are shown to have the low-access property. In particular, they have the optimal access property for repairing$u$failed nodes when the dimension of the code is divisible by the rack size$u$.

VCIP 2022 “Tire pattern image classification based on lightweight network challenge” aims to design lightweight networks that correctly classify tire surface tread patterns and indentation images using less overhead. To this end, we present a novel lightweight tire tread classification network. Concretely, we adopt the ShuffleNet-V2-x0.5 network as our backbone. To reduce the computation complexity, we introduce the Space-To-Depth and Anti-Alias Downsampling modules to pre-process the input image. Moreover, to enhance the classification ability of our model, we adopt the knowledge distillation strategy by considering Vision Transformer as the teacher network. To ensure the robustness of our model, we pre-train it on ImageNet and fine-tune the training set of the challenge. Experiments on the challenge dataset demonstrate that our model achieves supe-rior performance, with 99.00% classification accuracy, 25.51M FLOPs, and 0.20M parameters.

For image semantic segmentation, a fully convolutional network is usually employed as the encoder to abstract visual features of the input image. A meticulously designed decoder is used to decoding the final feature map of the backbone. The output resolution of backbones which are designed for image classification task is too low to match segmentation task. Most existing methods for obtaining the final high-resolution feature map can not fully utilize the information of different layers of the backbone. To adequately extract the information of a single layer, the multi-scale context information of different layers, and the global information of backbone, we present a new attention-augmented module named Dense-attention Context Module (DCM), which is used to connect the common backbones and the other decoding heads. The experiments show the promising results of our method on Cityscapes dataset.

The asymmetric cell division cycle of Caulobacter crescentus is orchestrated by an elaborate gene-protein regulatory network, centered on three major control proteins, DnaA, GcrA and CtrA. The regulatory network is cast into a quantitative computational model to investigate in a systematic fashion how these three proteins control the relevant genetic, biochemical and physiological properties of proliferating bacteria. Different controls for both swarmer and stalked cell cycles are represented in the mathematical scheme. The model is validated against observed phenotypes of wild-type cells and relevant mutants, and it predicts the phenotypes of novel mutants and of known mutants under novel experimental conditions. Because the cell cycle control proteins of Caulobacter are conserved across many species of alpha-proteobacteria, the model we are proposing here may be applicable to other genera of importance to agriculture and medicine (e.g., Rhizobium, Brucella).

Progression of a cell through the division cycle is tightly controlled at different steps to ensure the integrity of genome replication and partitioning to daughter cells. From published experimental evidence, we propose a molecular mechanism for control of the cell division cycle in Caulobacter crescentus. The mechanism, which is based on the synthesis and degradation of three "master regulator" proteins (CtrA, GcrA, and DnaA), is converted into a quantitative model, in order to study the temporal dynamics of these and other cell cycle proteins. The model accounts for important details of the physiology, biochemistry, and genetics of cell cycle control in stalked C. crescentus cell. It reproduces protein time courses in wild-type cells, mimics correctly the phenotypes of many mutant strains, and predicts the phenotypes of currently uncharacterized mutants. Since many of the proteins involved in regulating the cell cycle of C. crescentus are conserved among many genera of alpha-proteobacteria, the proposed mechanism may be applicable to other species of importance in agriculture and medicine.