Abstract: Local key action regions can improve CNN-based visual action recognition. Self-attention requires focusing on important details and ignoring others. Self-attention aids action recognition. Current self-attention methods ignore local feature vector correlations at spatial positions…
Abstract: Annotating chest X-rays saves time and effort. Recent weakly-supervised algorithms like multi-instance learning (MIL) and class activation maps (CAM) attempted this task, but they often produce inaccurate or incomplete regions. One reason is the…
Abstract: Long-term image restoration and manipulation goals include learning a good image prior. Existing methods like deep image prior (DIP) capture low-level image statistics, but there are still gaps toward capturing rich image semantics like…
Abstract: Internet of things sensors are enabling data-driven traffic speed prediction, a foundation of advanced traffic management. However, current research studies mostly predict traffic for one hour ahead. Long-term prediction methods have error accumulation, exposure…
Abstract: BCI technology connects the brain to external devices. P300 wave extraction from electroencephalography (EEG) recordings is a popular BCI protocol. P300 waves are event-related potentials with a 300-ms latency after rare stimuli. This paper…
Abstract: Vehicle detection, which identifies vehicles as axis-aligned bounding boxes in still images, is widely used to estimate autonomous vehicle (AV) range, time-to-collision, and motion. Bounding boxes are convenient but too coarse to adapt to…
Abstract: Conventional deformable registration methods solve an image pair optimization model, which is computationally expensive. Recent deep learning-based methods estimate deformation quickly. These data-driven heuristic network architectures lack geometric constraints needed to generate plausible topology-preserving…
Abstract: We propose a novel Convolutional Neural Network (CNN) architecture for joint detection and matching of feature points in images from different sensors using a single forward pass. Unlike classical approaches (SIFT, etc.), the feature…
Abstract: This study develops a high accident risk prediction model to analyze traffic accident data and prioritize intersection improvements. A traffic accident database was organized and analyzed, and an intersection accident risk prediction model based…
Abstract: Vision-based intelligent systems like driver assistance and transportation should consider weather. Haze in images can endanger drivers. Real-world haze density measures image visibility and usability. In outdoor vision-based intelligent systems, haze density prediction can…
Abstract: Many visual tasks use channel attention mechanisms to improve performance. It boosts informative channels and suppresses useless ones. Channel attention modules have been proposed and implemented recently. They rely on convolution and pooling. This…
Abstract: Existing vehicle re-identification methods use spatial pooling operations to aggregate feature maps extracted from off-the-shelf backbone networks like VGGNet, GoogLeNet, and ResNet. They neglect feature map spatial significance, lowering vehicle re-identification performance. First, an…
WhatsApp us