TUM: how to use TUM dataset - TUM数据集的使用

# 目标 - 如何使用TUM数据集？ - 如何用SLAM跑TUM数据集？ - 如何评价SLAM定位的好坏？ # 获取数据集 - [TUM RGBD dataset](https://vision.in.tum.de/data/datasets/rgbd-dataset/download) - [rgbd_benchmark](https://vision.in.tum.de/data/datasets/rgbd-dataset/tools#evaluation) For example: ```sh wget -c https://vision.in.tum.de/rgbd/dataset/freiburg3/rgbd_dataset_freiburg3_walking_halfsphere.tgz wget -c https://vision.in.tum.de/rgbd/dataset/freiburg3/rgbd_dataset_freiburg3_walking_rpy.tgz ``` Unzip: ```sh tar zxvf rgbd_dataset_freiburg3_walking_halfsphere.tgz ``` 数据集目录结构: ```sh ➜ rgbd_dataset_freiburg3_walking_rpy tree -L 1 . ├── accelerometer.txt ├── associations.txt ├── depth ├── depth.txt ├── groundtruth.txt ├── rgb └── rgb.txt ``` # 关联彩色图像与深度图像 先要下载RGBD benchmark ```sh associate.py rgb.txt depth.txt > associations.txt ``` associations.txt 内容格式: - RGB timestamp, float - RGB image, string - Depth timestamp, float - Depth image, string For example: ```sh 1341846647.802247 rgb/1341846647.802247.png 1341846647.802269 depth/1341846647.802269.png 1341846647.834093 rgb/1341846647.834093.png 1341846647.834105 depth/1341846647.834105.png ... ``` # 仿真 ## ORB SLAM2 运行参数： - Vocabulary - TUM.yaml - dataset directory - associations.txt For example: ```sh ./Examples/RGB-D/rgbd_tum ./Vocabulary/ORBvoc.txt ./Examples/RGB-D/TUM3.yaml /root/Dataset/TUM/freiburg3/rgbd_dataset_freiburg3_walking_rpy /root/Dataset/TUM/freiburg3/rgbd_dataset_freiburg3_walking_rpy/associations.txt ``` # 用数据集进行验证 - 先要下载 rgbd_benchmark_tools ```sh svn checkout https://svncvpr.in.tum.de/cvpr-ros-pkg/trunk/rgbd_benchmark/rgbd_benchmark_tools ``` - 下载 Ground truth ```sh wget -c https://vision.in.tum.de/rgbd/dataset/freiburg3/rgbd_dataset_freiburg3_walking_rpy-groundtruth.txt ``` - ATE Error of rgbd_dataset_freiburg3_walking_rpy ```sh evaluate_ate.py --plot ate.png --verbose --save_associations ate_associate.txt ../rgbd_dataset_freiburg3_walking_rpy-groundtruth.txt ../KeyFrameTrajectory.txt compared_pose_pairs 388 pairs absolute_translational_error.rmse 1.085461 m absolute_translational_error.mean 0.970126 m absolute_translational_error.median 1.135803 m absolute_translational_error.std 0.486909 m absolute_translational_error.min 0.047486 m absolute_translational_error.max 1.905858 m ```  - RPE Error of rgbd_dataset_freiburg3_walking_rpy ```sh evaluate_rpe.py --fixed_delta --plot rpe.png --save rpe.txt --verbose ../rgbd_dataset_freiburg3_walking_rpy-groundtruth.txt ../KeyFrameTrajectory.txt compared_pose_pairs 364 pairs translational_error.rmse 0.385744 m translational_error.mean 0.290880 m translational_error.median 0.203410 m translational_error.std 0.253352 m translational_error.min 0.000000 m translational_error.max 1.024328 m rotational_error.rmse 7.523788 deg rotational_error.mean 5.614611 deg rotational_error.median 0.064197 deg rotational_error.std 5.008347 deg rotational_error.min 0.000000 deg rotational_error.max 19.987115 deg ```  # groundtruth_file 数据格式 ```sh groundtruth_file ground-truth trajectory file (format: "timestamp tx ty tz qx qy qz qw") estimated_file estimated trajectory file (format: "timestamp tx ty tz qx qy qz qw") ``` For example: ```sh 1341846647.802247 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 1.0000000 1341846647.866652 0.0070123 0.0018382 -0.0068172 0.0032666 0.0019550 0.0018201 0.9999911 ... ``` # 2D到3D点的转换 ```cpp fx = 525.0 # focal length x fy = 525.0 # focal length y cx = 319.5 # optical center x cy = 239.5 # optical center y factor = 5000 # for the 16-bit PNG files # OR: factor = 1 # for the 32-bit float images in the ROS bag files for v in range(depth_image.height): for u in range(depth_image.width): Z = depth_image[v,u] / factor; X = (u - cx) * Z / fx; Y = (v - cy) * Z / fy; ``` # 相机内参 ```cpp Camera fx fy cx cy d0 d1 d2 d3 d4 (ROS default) 525.0 525.0 319.5 239.5 0.0 0.0 0.0 0.0 0.0 Freiburg 1 RGB 517.3 516.5 318.6 255.3 0.2624 -0.9531 -0.0054 0.0026 1.1633 Freiburg 2 RGB 520.9 521.0 325.1 249.7 0.2312 -0.7849 -0.0033 -0.0001 0.9172 Freiburg 3 RGB 535.4 539.2 320.1 247.6 0 0 0 0 0 ```