While CNN-based models have made remarkable progress on human pose estimation, what spatial dependencies they capture to localize keypoints remains unclear. In this work, we propose a model called Trans-Pose, which introduces Transformer for human pose estimation. The attention layers built in Transformer enable our model to capture long-range relationships efficiently and also can reveal what dependencies the predicted key-points rely on. To predict keypoint heatmaps, the last attention layer acts as an aggregator, which collects contributions from image clues and forms maximum positions of keypoints. Such a heatmap-based localization approach via Transformer conforms to the principle of Activation Maximization [19]. And the revealed dependencies are image-specific and fine-grained, which also can provide evidence of how the model handles special cases, e.g., occlusion. The experiments show that TransPose achieves 75.8 AP and 75.0 AP on COCO validation and test-dev sets, while being more lightweight and faster than mainstream CNN architectures. The TransPose model also transfers very well on MPII benchmark, achieving superior performance on the test set when fine-tuned with small training costs. Code and pre-trained models are publicly available 1 .
Aircraft four dimensional (4D, including longitude, latitude, altitude and time) trajectory prediction is a key technology for existing automation systems and the basis for future trajectory-based operations. This paper firstly summarizes the background and significance of the trajectory prediction problems and then introduces the definition and basic process of trajectory prediction, including four modules: preparation, prediction, update, and output. In addition, the trajectory prediction methods are summarized into three types: the state estimation model, the Kinetic model, and the machine learning model, and in-depth analysis of various models is carried out. Further, the relevant databases required for the study are introduced, including the aircraft performance database, aircraft monitoring database, and meteorological database. Finally, challenges and future development directions of the current trajectory prediction problem are summarized.
Current state-of-the-art trajectory methods do not perform well in the terminal airspace that surrounds an airport due to its complex airspace structure and the frequently changing flight postures of aircraft. Since an aircraft that takes off or lands in an airport must follow a specified procedure, this paper will learn a data-driven trajectory prediction model from many historical trajectories to improve the accuracy and robustness of trajectory prediction in the terminal airspace. A regularization method is utilized to reconstruct each aircraft trajectory to obtain a high-quality trajectory with equal time intervals and no noise. Furthermore, we formulate the 4D trajectory prediction problem as a sequence-to-sequence learning problem, and we propose a sequence-to-sequence deep long short-term memory network (SS-DLSTM) for trajectory prediction, which can effectively capture the long and short temporal dependencies and the repetitive nature among trajectories. The proposed model is composed of an encoding module and a decoding module, where the encoding mode realizes the feature representation of historical trajectories, while the decoding module accepts the output of the encoding module as its initial input and recursively outputs the predicted trajectory sequence. The proposed method is applied to a dataset for the terminal airspace in Guangzhou, China. The experimental results demonstrate that our approach has relatively high robustness and outperforms mainstream data-driven trajectory prediction methods in terms of accuracy.
3 May 2019 in IEEE Transactions on Neural Networks and Learning Systems
Learning long-term dependences (LTDs) with recurrent neural networks (RNNs) is challenging due to their limited internal memories. In this paper, we propose a new external memory architecture for RNNs called an external addressable long-term and working memory (EALWM)-augmented RNN. This architecture has two distinct advantages over existing neural external memory architectures, namely the division of the external memory into two parts--long-term memory and working memory--with both addressable and the capability to learn LTDs without suffering from vanishing gradients with necessary assumptions. The experimental results on algorithm learning, language modeling, and question answering demonstrate that the proposed neural memory architecture is promising for practical applications.
24 March 2021 in Journal of Advanced Transportation
Due to the strong propagation causality of delays between airports, this paper proposes a delay prediction model based on a deep graph neural network to study delay prediction from the perspective of an airport network. We regard airports as nodes of a graph network and use a directed graph network to construct airports’ relationship. For adjacent airports, weights of edges are measured by the spherical distance between them, while the number of flight pairs between them is utilized for airports connected by flights. On this basis, a diffusion convolution kernel is constructed to capture characteristics of delay propagation between airports, and it is further integrated into the sequence-to-sequence LSTM neural network to establish a deep learning framework for delay prediction. We name this model as deep graph-embedded LSTM (DGLSTM). To verify the model’s effectiveness and superiority, we utilize the historical delay data of 325 airports in the United States from 2015 to 2018 as the model training set and test set. The experimental results suggest that the proposed method is superior to the existing mainstream methods in terms of accuracy and robustness.
About Me
My research interests include natural language processing, computer vision, and robotic learning.
My research interests include natural language processing, computer vision, and robotic learning.