3D human action recognition is crucial in broad industrial application scenarios such as robotics, video surveillance, autonomous driving, or intellectual education, etc. In this paper, we present a new point cloud sequence network called PointMapNet for 3D human action recognition. In PointMapNet, two point cloud feature maps symmetrical to depth feature maps are proposed to summarize appearance and motion representations from point cloud sequences. Specifically, we first convert the point cloud frames to virtual action frames using static point cloud techniques. The virtual action frame is a 1D vector used to characterize the structural details in the point cloud frame. Then, inspired by feature map-based human action recognition on depth sequences, two point cloud feature maps are symmetrically constructed to recognize human action from the point cloud sequence, i.e., Point Cloud Appearance Map (PCAM) and Point Cloud Motion Map (PCMM). To construct PCAM, an MLP-like network architecture is designed and used to capture the spatio-temporal appearance feature of the human action in a virtual action sequence. To construct PCMM, the MLP-like network architecture is used to capture the motion feature of the human action in a virtual action difference sequence. Finally, the two point cloud feature map descriptors are concatenated and fed to a fully connected classifier for human action recognition. In order to evaluate the performance of the proposed approach, extensive experiments are conducted. The proposed method achieves impressive results on three benchmark datasets, namely NTU RGB+D 60 (89.4% cross-subject and 96.7% cross-view), UTD-MHAD (91.61%), and MSR Action3D (91.91%). The experimental results outperform existing state-of-the-art point cloud sequence classification networks, demonstrating the effectiveness of our method.

Cervical cancer, the fourth most common fatal cancer in women, has a considerably increased cure rate if identified and treated at early stages. Computer-aided diagnosis technology is the key for scaling up cervical cancer screening. Computer-processed images of nuclei can help doctors better analyze and diagnose the extent of cancer lesions. However, the lesion characteristics of cervical cancer are mainly reflected in the cell nuclei, which will be significantly larger, with distorted boundary shapes and deepened colors. What's worse is that the nuclei of diseased cells may be hidden in cell clusters, making it difficult for existing medical image segmentation models to achieve satisfactory results. In this paper, based on Deeplabv3+, we propose a HR-ASPP model with accurate spatial localization and better shape feature extraction to segment cervical nuclei. Firstly, HR-ASPP uses HRNet as the backbone network, which has good feature extraction ability for small targets. The structure of parallel multi-resolution and repeated multi-scale fusion makes the target information avoid losing in the deep network, so as to better extract the location information of cell nuclei. Secondly, we devise a lighter weight deformable convolution. The proposed HR-ASPP replaces part of the convolution in HRNet with a lighter weight deformable convolution to learn the edge shape of aberrant nuclei better while reducing the network computation. In addition, HR-ASPP incorporate CARAFE, which is able to aggregate information within a large receptive field and dynamically adapt to the content of a particular instance. The HR-ASPP method proposed in this paper achieves about 50.64%, 84.98% intersection over union (IoU) respectively on our dataset and the public datasets-ISBI. The results indicate that the proposed method achieves the best performance when compared with state-of-the-art approaches for pathologic image segmentation.

Accurate vessel trajectory prediction is vital for forecasting navigation trends, reducing potential risks, and ensuring maritime traffic safety. Using automatic identification system (AIS) data to improve the comprehensive performance of vessel future trajectory prediction is an urgent problem in intelligent transportation systems. In this paper, an improved Pelican optimization algorithm (IPOA) optimized bidirectional gated recurrent unit-attention mechanism (BiGRU-AM) vessel trajectory prediction framework based on AIS data is proposed. This method integrates the optimized hyperparameters of IPOA into the BiGRU-AM backbone network. Utilizing the BiGRU-AM network with tuned parameters to achieve bidirectional information enhancement, and establish an efficient IPOA-BiGRU-AM hybrid prediction model. Furthermore, we propose a shared fusion training mechanism (SFTM) for information exchange between modules. A comprehensive evaluation system covering global error, local error and spatial similarity is developed to address the limitations of single indicator evaluation. The trajectory prediction effectiveness of this method is verified on three datasets. Compared with the current methods, this approach improves accuracy by up to 79.30% across all indicators and reduces the average prediction error by 76.31%. In terms of model complexity, our model’s running speed has improved by 49.73% on average, and has a 1.5 times computational advantage compared to baseline models. The experimental results proved the superiority of the proposed method.

Plant disease classification is critical for effective crop management. Recent advances in deep learning, especially Vision Transformers (ViTs), have shown promise due to their strong global feature modeling capabilities. However, ViTs often overlook local features and suffer from feature extraction degradation during patch merging as channels increase. To address these issues, we propose PLTransformer, a hybrid model designed to symmetrically capture both global and local features. We design a symmetric multi-scale convolutional module that combines two different-scale receptive fields to simultaneously extract global and local features so that the model can better perceive multi-scale disease morphologies. Additionally, we propose an overlap-attentive channel downsampler that utilizes inter-channel attention mechanisms during spatial downsampling, effectively preserving local structural information and mitigating semantic loss caused by feature compression. On the PlantVillage dataset, PLTransformer achieves 99.95% accuracy, outperforming DeiT (96.33%), Twins (98.92%), and DilateFormer (98.84%). These results demonstrate its superiority in handling multi-scale disease features.