The irregular domain and lack of ordering make it challenging to design deep neural networks for point cloud processing. This paper presents a novel framework named Point Cloud Transformer (PCT) for point cloud learning. PCT is based on Transformer, which achieves huge success in natural language processing and displays great potential in image processing. It is inherently permutation invariant for processing a sequence of points, making it well-suited for point cloud learning. To better capture local context within the point cloud, we enhance input embedding with the support of farthest point sampling and nearest neighbor search. Extensive experiments demonstrate that the PCT achieves the state-of-the-art performance on shape classification, part segmentation, semantic segmentation, and normal estimation tasks.

Humans can naturally and effectively find salient regions in complex scenes. Motivated by this observation, attention mechanisms were introduced into computer vision with the aim of imitating this aspect of the human visual system. Such an attention mechanism can be regarded as a dynamic weight adjustment process based on features of the input image. Attention mechanisms have achieved great success in many visual tasks, including image classification, object detection, semantic segmentation, video understanding, image generation, 3D vision, multimodal tasks, and self-supervised learning. In this survey, we provide a comprehensive review of various attention mechanisms in computer vision and categorize them according to approach, such as channel attention, spatial attention, temporal attention, and branch attention; a related repository https://github.com/MenghaoGuo/Awesome-Vision-Attentions is dedicated to collecting related work. We also suggest future directions for attention mechanism research.

No abstract

We measure the stability and folding rate of a mutant of the enzyme phosphoglycerate kinase (PGK) inside bone tissue cells as a function of temperature from 38 to 48°C. To facilitate measurement in individual living cells, we developed a rapid laser temperature stepping method capable of measuring complete thermal melts and kinetic traces in about two min. We find that this method yields improved thermal melts compared to heating a sample chamber or microscope stage. By comparing results for six cells with in vitro data, we show that the protein is stabilized by about 6 kJ∕mole in the cytoplasm, but the temperature dependence of folding kinetics is similar to in vitro. The main difference is a slightly steeper temperature dependence of the folding rate in some cells that can be rationalized in terms of temperature-dependent crowding, local viscosity, or hydrophobicity. The observed rate coefficients can be fitted within measurement uncertainty by an effective two-state model, even though PGK folds by a multistate mechanism. We validate the effective two-state model with a three-state free energy landscape of PGK to illustrate that the effective fitting parameters can represent a more complex underlying free energy landscape. A s the ability of molecular dynamics and physics-based force fields to simulate protein folding kinetics improves (1-3), folding kinetics in complex environments such as crowding agents (4-6) or living cells (7) is receiving increased attention. Energy landscape theory predicts that the free energies of different protein states are similar, and that the free energy barriers connecting these states are small (8); hence, one might expect that even small perturbations in vivo (on the order of RT ≈ 2.5 kJ∕mole) can have significant effects on folding and function: the relative population of two pathways is exponentially sensitive to their free energy difference (Boltzmann factor). In this manner, different microenvironments in the cell could exert localized "postpost-translational" control over protein structure, folding, and function (9).Due to their dynamic nature and small folding equilibrium constant, proteins spontaneously unfold and refold many times in vivo between translational synthesis and degradation. We recently developed Fast Relaxation Imaging (FReI) to study posttranslational unfolding and refolding kinetics inside cells (10). The technique applies small temperature upward or downward jumps to living cells and monitors a target protein that has been FRET-labeled to distinguish native and unfolded states by fluorescence microscopy. As our prototype protein, we chose a FRETlabeled phosphoglycerate kinase construct (FRET-PGK) for comparison between in cell and in vitro. This multistate folder should be particularly sensitive to cellular modulation of its folding rate and mechanism. We found that unfolding was highly reversible in the cytoplasm up to 43°C (10). The local viscosity experienced by the polypeptide chain is about twice as large as in aqueous solvent (7). Significant kine...

Automated brain lesion segmentation provides valuable information for the analysis and intervention of patients. In particular, methods that are based on convolutional neural networks (CNNs) have achieved state-of-the-art segmentation performance. However, CNNs usually require a decent amount of annotated data, which may be costly and time-consuming to obtain. Since unannotated data is generally abundant, it is desirable to use unannotated data to improve the segmentation performance for CNNs when limited annotated data is available. In this work, we propose a semi-supervised learning (SSL) approach to brain lesion segmentation, where unannotated data is incorporated into the training of CNNs. We adapt the mean teacher model, which is originally developed for SSL-based image classification, for brain lesion segmentation. Assuming that the network should produce consistent outputs for similar inputs, a loss of segmentation consistency is designed and integrated into a self-ensembling framework. Self-ensembling exploits the information in the intermediate training steps, and the ensemble prediction based on the information can be closer to the correct result than the single latest model. To exploit such information, we build a student model and a teacher model, which share the same CNN architecture for segmentation. The student and teacher models are updated alternately. At each step, the student model learns from the teacher model by minimizing the weighted sum of the segmentation loss computed from annotated data and the segmentation consistency loss between the teacher and student models computed from unannotated data. Then, the teacher model is updated by combining the updated student model with the historical information of teacher models using an exponential moving average strategy. For demonstration, the proposed approach was evaluated on ischemic stroke lesion segmentation. Results indicate that the proposed method improves stroke lesion segmentation with the incorporation of unannotated data and outperforms competing SSL-based methods.