【论文】室内定位数据集Navvis

Image-based localization using LSTMs for structured feature correlation

Abstract

  • CNN+LSTM architecture
    • CNN:鲁棒性,模糊/照明变化,学习定位特征
    • LSTM:CNN输出,特征向量降维
  • camera pose regression
  • indoor and outdoor scenes

Introduction

  • 目标:从图像 - 估计相机pose - 定位设备/车辆
  • 应用:导航、自动驾驶、移动机器人、增强现实、运动结构sfm
  • 对比:
    • SIFT[37],方法[33,46,52,64]
      • 特征匹配:查询图像特征2D-SfM模型3D点
      • 姿势估计:n-point solver(RANSAC loop)
    • CNN方法[25,26]
  • 端到端定位:
    • 分类问题:PlaNet 适用大环境,方向准确性受制于训练样本
    • 回归问题:PoseNet 6DoF

Contribution

  • CNN:鲁棒性,模糊/照明变化,学习定位特征
    • PoseNet:FC层->高纬特征->回归pose 不是最佳的、过拟合,dropout策略
  • LSTM:FC层-LSTM,降维,选择相关性特征

Dataset

  • laser scanner(accurate ground truth)
  • TUM-LSI indoor dataset :重复结构、弱纹理表面、提供真值
  • 1314张,5575㎡
  • ID: 2015-08-16_15.34.11
  • 摄像头:camera0-camera4 (5个)
  • 像素:旋转->缩放256->裁剪224*224
  • 其他数据集:[49]
  • 下载
  • 分类:
    • images / cam0-cam4
    • poses / xml
    • geo-refrence.xml
    • train / test .txt

Further reading

  • [25] A. Kendall and R. Cipolla. Modelling uncertainty in deep learning for camera relocalization. In IEEE International Conference on Robotics and Automation (ICRA), 2016. 1, 2,3, 4, 5, 6, 7
  • [26] A. Kendall, M. Grimes, and R. Cipolla. Posenet: A convolutional network for real-time 6-dof camera relocalization. In IEEE International Conference on Computer Vision (ICCV),2015.
  • [37] D. Lowe. Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision (IJCV),2004.
  • [33] Y. Li, N. Snavely, D. Huttenlocher, and P. Fua. Worldwide pose estimation using 3d point clouds. In European Conference on Computer Vision (ECCV), 2012. 1, 2, 6
  • [46] T. Sattler, B. Leibe, and L. Kobbelt. Efficient & Effective Prioritized Matching for Large-Scale Image-Based Localization. IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 2016 (to appear)
  • [52] L. Sv ̈arm, O. Enqvist, F. Kahl, and M. Oskarsson. City-Scale Localization for Cameras with Known Vertical Direction.IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 2016 (to appear)
  • [64] B. Zeisl, T. Sattler, and M. Pollefeys. Camera pose voting for large-scale image-based localization. In IEEE International Conference on Computer Vision (ICCV), 2015.
  • [49] J. Shotton, B. Glocker, C. Zach, S. Izadi, A. Criminisi, and A. Fitzgibbon. Scene coordinate regression forests for camera relocalization in rgb-d images. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013.
发布于 2019-10-30 03:40