Abstract: ReID is difficult due to person images’ large pose variations and misalignment errors. Existing works use pose estimation, part segmentation, etc. to improve pedestrian representations. Boosting ReID accuracy requires computational overheads and makes deep…
Abstract: Road intersections have always caused traffic congestion, but an efficient transportation system can benefit society. Real-time traffic signal timing may reduce such traffic congestion. Most traffic signal control methods require a lot of road…
Abstract: Autonomous Driving’s most important module is pedestrian detection. Cameras are often used for this, but their quality suffers in low-light nighttime driving. A thermal camera’s image quality is unaffected in similar conditions. This paper…
Abstract: Vehicle re-identification (Re-ID) research has grown due to intelligent traffic control and monitoring. Vehicle Re-ID is more difficult and unpredictable than person Re-ID because viewpoint variations greatly affect vehicle appearance. Existing studies mostly extract…
Abstract: Transport mode detection is a classification problem to design an algorithm to infer a user’s transport mode from multimodal signals (GPS and/or inertial sensors). It can track carbon footprints, analyze mobility behavior, and plan…
Abstract: Deep learning has improved short-term traffic forecasting. However, traffic flow stochasticity and distribution imbalance cause uncertainty and network overfitting during deep learning. This paper proposes M-B-LSTM, an end-to-end hybrid deep learning network model for…
Abstract: Many image understanding tasks require natural scene text spotting. Text detection and recognition are subtasks. We propose a unified network that localizes and recognizes text in a single forward pass without image cropping, feature…
Abstract: Understanding human-object interaction (HOI) requires detection. Interactiveness Knowledge—whether humans and objects interact—is examined in this paper. Interactiveness knowledge can be learned across HOI datasets and bridge diverse HOI category settings. We use an Interactiveness…
Abstract: Rainy days with ponding water on the road surface can cause vehicle crashes, injuries, and deaths. Surveillance video-based splashed water detection is a promising way to reduce traffic accidents. Surveillance videos have many lighting…
Abstract: Protein sequence data from computational biology and bioinformatics is ideal for NLP Language Models (LMs). Low-inference LMs seek new prediction frontiers. We trained two auto-regressive models (Transformer-XL, XLNet) and four auto-encoder models (BERT, Albert,…
Abstract: Automatic vessel segmentation in fundus images helps screen, diagnose, treat, and evaluate cardiovascular and ophthalmologic diseases. Retinal vessel segmentation has been difficult due to limited well-annotated data, varying vessel sizes, and intricate vessel structures….
Abstract: Deep neural networks succeed because of good architectures. Some Neural Architecture Search (NAS) methods search or manually design deep architectures. Even a well-designed/searched architecture may contain many nonsignificant or redundant modules/operations (e.g., intermediate convolution…
WhatsApp us