Metro Origin-Destination (OD) prediction is a crucial yet challenging spatial-temporal prediction task in urban computing, which aims to accurately forecast cross-station ridership for optimizing metro scheduling and ...
详细信息
Generative adversarial networks (GANs) are known for their strong abilities on capturing the underlying distribution of training instances. Since the seminal work of GAN, many variants of GAN have been proposed. Howev...
详细信息
Package delivery via the UAVs is a promising transport mode to provide efficient and green logistic services, especially in urban areas or complicated topography. However, the energy storage limit of the UAV makes it ...
详细信息
The existence of pulse radio frequency interference (PRFI) severely affects the signal processing and image interpretation of synthetic aperture radar (SAR), which cannot be ignored. In SAR dataprocessing, the detect...
详细信息
The existence of pulse radio frequency interference (PRFI) severely affects the signal processing and image interpretation of synthetic aperture radar (SAR), which cannot be ignored. In SAR dataprocessing, the detection accuracy directly affects the performance of the PRFI mitigation algorithm. However, most of the existing detectors can only detect the occurrence of PRFI, but cannot locate the PRFI. Based on a two-step process combining eigenvalue decomposition (EVD) with short-time Fourier transform (STFT), in this paper, a novel PRFI detector, which has the capability to detecting and locating PRFI, is proposed. In the proposed detector, the first step is the detection of PRFI occurrence, which is based on the Hankel structure and EVD. The second step is the location of PRFI, which is based on the STFT and a simple ratio operator. Experimental results, which are based on simulated and measured SAR data, validate the performance of the proposed PRFI detector.
Sentiment analysis is an important and challenging task in the field of natural language processing. Researchers have used vocabulary-based methods and machine learning methods to conduct research on sentiment analysi...
详细信息
Sentiment analysis is an important and challenging task in the field of natural language processing. Researchers have used vocabulary-based methods and machine learning methods to conduct research on sentiment analysis tasks. In recent years, deep learning has achieved great success in the field of sentiment analysis. However, for complex textual data, using a single model is often insufficient. Inspired by deep learning models, in this study, we proposed a hybrid model using BiLSTM and attention mechanism, called BiLSTM-ATT model, to solve the problem of sentiment analysis. First, we adopted the GloVe method to train the initialized word embeddings. GloVe converts textual information into word vectors, which can calculate the distance between words. Next, we used a convolutional neural network that can extract local features, and a BiLSTM that can extract long-range semantic features of text with bidirectional extraction of long-term dependencies. Finally, the attention mechanism was used to improve the performance of the model by calculating the weight of the data. Experimental results demonstrate that our proposed hybrid BiLSTM-ATT model outperforms traditional deep learning methods in Accuracy, Recall, and Fl-Score. Our method was compared with deep learning methods on the IMDB movie review dataset.
Recently, there has been a surge in face personalization techniques, benefiting from the advanced capabilities of pretrained text-to-image diffusion models. Among these, a notable method is Textual Inversion, which ge...
详细信息
Different from conventional federated learning, personalized federated learning (PFL) is able to train a customized model for each individual client according to its unique requirement. The mainstream approach is to a...
详细信息
In recent years, oriented object detection technology has made significant progress. Most oriented object detectors represent oriented bounding boxes by adding additional angle information to the horizontal bounding b...
In recent years, oriented object detection technology has made significant progress. Most oriented object detectors represent oriented bounding boxes by adding additional angle information to the horizontal bounding box representation, which leads to serious angle periodicity issues. To address these issues, this paper proposes double midpoint vectors (DMvectors) to represent oriented objects. For any oriented bounding box, we align the center point with the origin in the Cartesian coordinate system, and evenly distribute the midpoints of the box edges across the four quadrants. We use the midpoint of the first and second quadrants, as well as the width and height of the box, to represent the oriented bounding box. DMvectors solve the periodic problem of angles by implicitly including angle information in vectors. To demonstrate the effectiveness of our proposed method, we conducted extensive experiments on two common remote sensing datasets (DOTA and HRSC2016), and our method achieved competitive results.
Federated learning (FL) is a prospective distributed machine learning framework that can preserve data privacy. In particular, cross-silo FL can complete model training by making isolated data islands of different org...
详细信息
This paper introduces the Temporal Transform Network based on Scale Sequences (TTS) for cloth-changing person re-identification in video datasets. The TTS network is designed to capture multi-scale temporal cues withi...
This paper introduces the Temporal Transform Network based on Scale Sequences (TTS) for cloth-changing person re-identification in video datasets. The TTS network is designed to capture multi-scale temporal cues within video sequences. It accomplishes this by initially modeling short-term temporal cues between adjacent frames, followed by capturing long-term relationships between non-consecutive frames. In more detail, short-term temporal cues are modeled through parallel inflated convolutions with different time dilation rates, enabling the representation of pedestrian movement and appearance dynamics. Long-term relationships are effectively captured using a temporal self-attention model, mitigating challenges such as occlusion and noise within the video sequence. The TTS network outperforms existing methods across cloth-changing video ReID datasets such as CCVID. For instance, under general settings, our approach exhibits a 1.1% improvement in top-1 accuracy and a corresponding 1.1% increase in mAP compared to the baseline. In cloth-changing settings, we observe a 0.2% enhancement in top-1 accuracy and a notable 1.3% increase in mAP relative to the baseline.
暂无评论