In this paper we propose an end-to-end learnable approach that detects static urban objects from multiple views, re-identifies instances, and finally assigns a geographic position per object. Our method relies on a Graph Neural Network (GNN) to, detect all objects and output their geographic positions given images and approximate camera poses as input. Our GNN simultaneously models relative pose and image evidence, and is further able to deal with an arbitrary number of input views. Our method is robust to occlusion, with similar appearance of neighboring objects, and severe changes in viewpoints by jointly reasoning about visual image appearance and relative pose. Experimental evaluation on two challenging, large-scale datasets and comparison with state-of-the-art methods show significant and systematic improvements both in accuracy and efficiency, with 2-6% gain in detection and re-ID average precision as well as 8x reduction of training time.
Clutt3R-Seg: Sparse-view 3D Instance Segmentation for Language-grounded Grasping in Cluttered Scenes
Dept. of Mechanical Engineering, SNU(机械工程系,首尔国立大学);Interdisciplinary Program in Artificial Intelligence, SNU(人工智能跨学科项目,首尔国立大学);Robotics Lab, Hyundai Motor Company(Hyundai Motor Company机器人实验室)
Spatially-Weighted CLIP for Street-View Geo-localization
School of Geospatial Engineering and Science, Sun Yat-Sen University(中山大学地理科学与规划学院);School of Geographical and Earth Sciences, University of Glasgow(格拉斯哥大学地理与地球科学学院);School of System Science and Engineering, Sun Yat-Sen University(中山大学系统科学与工程学院)
HumanRecon: Neural Reconstruction of Dynamic Human Using Geometric Cues and Physical Priors
Beijing University of Posts and Telecommunications(北京邮电大学);DJI Technology(大疆创新科技);Zhejiang University(浙江大学);Tampere Universities(坦佩雷大学);University of Adelaide(阿德莱德大学)
Multi-view Surface Reconstruction Using Normal and Reflectance Cues
University of Zurich(苏黎世大学);Université de Toulouse(图卢兹大学);CNRS, UNICAEN, ENSICAEN, Normandie Université(法国国家科学研究中心、UNICAEN、ENSICAEN、诺曼底大学);FittingBox(FittingBox公司);University of Copenhagen(哥本哈根大学)
GeoBridge++: Fact-Guided Geo-Semantic Bridging for Unified Cross-View Geo-Localization
School of Computer Science and Technology, Jilin University(吉林大学计算机科学与技术学院);School of Computer Science, Wuhan University(武汉大学计算机学院);Zhongguancun Academy, Beijing, China(中关村学院,北京,中国);State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University(武汉大学测绘遥感信息工程国家重点实验室)
Aerial Multi-View Stereo via Adaptive Depth Range Inference and Normal Cues
Department of Information Science and Technology, Ocean University of China(信息科学与技术学院,中国海洋大学);School of Computing and Mathematical Sciences, University of Leicester(计算与数学科学学院,莱斯特大学);Department of Electrical and Computer Engineering, Mississippi State University(电气与计算机工程学院,密苏里州立大学)
Beyond Frontal Faces: Improving Person Recognition Using Multiple Cues
UC Berkeley(加州大学伯克利分校);Facebook AI Research(Facebook人工智能研究院)
我们探讨了在一个不受约束的设置中识别相册中的人的身份的任务。为了促进这一点,我们介绍了新的人在相册(PIPA)数据集,包括超过60000个实例的2000个人从公共Flickr相册收集。由于只有大约一半的人的图像包含正面的脸,识别任务是非常具有挑战性的,由于姿势,服装,相机的观点,图像分辨率和照明的大变化。我们提出了Pose Invariant PErson Recognition(PIPER)方法,该方法积累了由深度卷积网络训练的poselet级别的人识别器的线索,以折扣姿态变化,结合面部识别器和全局识别器。在三种不同设置上的实验证实,在我们的无约束设置中,PIPER显着提高了DeepFace的性能,DeepFace是在LFW数据集上测量的最佳人脸识别器之一。
PlaNet - Photo Geolocation with Convolutional Neural Networks
Google(谷歌);RWTH Aachen University(亚琛工业大学)
有没有可能建立一个系统来确定一张照片的拍摄位置,只使用它的像素?一般来说,这个问题似乎异常困难:构建无法推断位置的情况是微不足道的。然而,图像通常包含信息线索,如地标,天气模式,植被,道路标记和建筑细节,结合起来可以让人们确定一个近似的位置,偶尔一个确切的位置。GeoGuessr和View from your Window等网站表明,人类相对擅长将这些线索整合到地理定位图像中,尤其是在地图上。在计算机视觉中,照片定位问题通常使用图像检索方法来处理。相比之下,我们将地球表面细分为数千个多尺度地理单元,并使用数百万个地理标记图像训练深度网络,从而将问题视为分类问题。虽然以前的方法只能识别地标或使用全局图像描述符进行近似匹配,但我们的模型能够使用和整合多个可见线索。我们发现,由此产生的模型,称为PlaNet,优于以前的方法,甚至在某些情况下达到超人水平的准确性。此外,我们通过将其与长短期记忆(LSTM)架构相结合,将我们的模型扩展到相册。通过学习利用时间相干性对不确定的照片进行地理定位,我们证明了该模型比单图像模型的性能提高了50%。
MonSter++: Unified Stereo Matching, Multi-view Stereo, and Real-time Stereo with Monodepth Priors
School of Electronic Information and Communications, Huazhong University of Science and Technology(电子信息与通讯学院,华中科技大学);Meta;AI Chip Center, Hong Kong University of Science and Technology(香港科技大学人工智能芯片中心);Nanjing Forestry University(南京林业大学);Autel Robotics
View-Consistent 3D Scene Editing via Dual-Path Structural Correspondense and Semantic Continuity
Wangxuan Institute of Computer Technology, Peking University(北京大学王萱计算机技术研究所);Department of Computer Science and Engineering, The Hong Kong University of Science and Technology(香港科技大学计算机科学与工程系)
Towards Interpretable Geo-localization: a Concept-Aware Global Image-GPS Alignment Framework
Peking University(北京大学);Harbin Institute of Technology(哈尔滨工业大学);The Hong Kong University of Science and Technology(香港科技大学);Harbin Institute of Technology (Weihai)(哈尔滨工业大学(威海))
Crafting Monocular Cues and Velocity Guidance for Self-Supervised Multi-Frame Depth Learning
Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Artificial Intelligence, University of Chinese Academy of Science(中国科学院大学人工智能学院);PhiGent Robotics(鉴智机器人);Southeast University(东南大学)