Reporter: bet365 online sports betting live University bet365 online sports betting live
Reporting location:#Tencent Conference:943-963-794
bet365 online sports betting live time:2022Year11month1Sunday (Tuesday) morning10Point
bet365 online sports betting live title: Locating and Counting Heads in Crowds With a Depth Prior
Personal introduction: Gao Shenghua, researcher at bet365 online sports betting live University(Tenured Professor)，Selected into the Youth Project of the National Overseas High-level bet365 online sports betting live Program，Shanghai Pujiang bet365 online sports betting live Plan，Dawn Scholar, an outstanding bet365 online sports betting live leader in Shanghai. bet365 online sports betting live interests include image and video processing and understanding, 3D reconstruction and image and video content editing and generation。2008Graduated from the bet365 online sports betting live of Science and Technology of China in 2015.2012Graduated from Nanyang Technological bet365 online sports betting live, Singapore。Later worked as a research scientist at the Institute for Advanced Study, bet365 online sports betting live of Illinois, Singapore。2014Joined the School of Information, bet365 online sports betting live University。so far，Published papers in top conferences and journals in the field of computer vision and artificial intelligence120More articles, total citations: nearly10000times. He served more than ten timesICCV/CVPR/AAAIField chair of other bet365 online sports betting live，Top journals in the field of computer visionIEEE TCSVTandNeurocomputing's deputy editor, etc.，Served as chairman of bet365 online sports betting live and seminars for more than ten times。
bet365 online sports betting live Summary: We resort to detection-based crowd counting by leveraging RGB-D data and design a dual-path guided detection network (DPDNet). Specifically, we propose bet365 online sports betting live guided detection module, which leverages density map to improve the head/non-head classification in detection network where the density implies the probability of a pixel being a head, and a depth-adaptive kernel that considers the variances in head sizes is also introduced to generate high-fidelity density map for more robust density map regression. We utilize such bet365 online sports betting live for post-processing of head detection and propose bet365 online sports betting live guided NMS strategy. Meanwhile, we also propose a depth-guided detection module to generate a dynamic dilated convolution to extract features of heads of different scales, and a depth-aware anchor. Then we use the bounding boxes whose sizes are generated with depth to train our DPDNet. We collect two large-scale RGB-D crowd counting datasets, which comprises a synthetic dataset and a real-world dataset, respectively. Since the depth value at long-distance positions cannot be obtained in the real-world dataset, we further propose a depth completion method with meta learning. Extensive experiments show that our method achieves the best performance for RGB-D crowd counting and localization.
