the reference answer GPT ᴼ ᴾ ᴱ ᴺ ᴬ ᴵ < br / > Colmap is a based on structured light, RGB - D, multiple points of view such as sensor data of 3 D reconstruction tools, can handle different types of data sets, And output the corresponding camera position, point cloud information. For autopilot datasets, it is generally necessary to do some pre-processing on the dataset in order to output data in the same format as colmap.
Here are some possible approaches:
-
Using SfM algorithm to preprocess data sets
Structure from Motion(SfM) algorithm similar to colmap, such as OpenSfM and OpenMVG, can be used to preprocess automatic driving data sets. These algorithms usually require input of RGB image sequence and IMU data, and output of camera pose, 3D point cloud and other information. You can use the preprocessed output to compare with the colmap output to verify that the data format is the same. -
Manually Annotate camera pose
If sensor data, such as GPS and IMU, has been provided in the data set, manually annotate camera pose. Specifically, you can use sensor data to calculate camera pose and then save the pose information in the same format as colmap. This approach can be time-consuming, but it avoids the problems that can occur when using the SfM algorithm. -
Write your own data handler
If neither of the above methods work, you can write your own data handler to convert the autopilot data set into the same format as colmap. Specifically, you can first understand the details of the colmap output format and then, for the autopilot data set, write a program to convert the data to the appropriate format. This approach requires some programming ability and understanding of the data format, but ensures that the data format is the same as colmap.
In general, processing the autopilot data set to obtain the same output format as colmap requires a certain amount of technology and time investment, but it can effectively make different data sets can be used for the same research problem.
