Differences

This shows you the differences between two versions of the page.

--- add_new_algorithm_and_dataset [2025/03/31 03:31] – created liuxzh12023
+++ add_new_algorithm_and_dataset [2025/04/12 17:00] (current) – liuxzh12023
@@ Line 1: / Line 1: @@
-Here we give NDT-LOAM as a example to show how we make a new algorithm as a docker images that can be used for SLAM-Hive.
+Here we give NDT-LOAM as a example to show how we make a new algorithm as a docker images that can be used for SLAM-Hive. (And it's also a good way for you to follow exist algorithm and dataset folder to build a same structure of your new one.)
+== Algorithm ==
 We first git clone the repo on github, and check how to start this algorithm in a normal way.
 {{:ndt1.png?1400|}}
+You need to create a folder in slam_hive_algos : NDT-LOAM, and clone the repo here. In this folder, you need also create slamhive folder, Dockerfile following: https://github.com/SLAM-Hive/slam_hive_web/blob/main/repo_overview.md
+Then according its readme, we build the slamhive folder. We find the parameters the NDT-LOAM provided and write them to yaml file and the SLAM-Hive parameter template. And the create the mapping.py file.
+{{:ndt2.png?1400|}}
+Then we create Configurations for NDT-LOAM, selecting the parameter template we created before.
+{{:ndt3.png?1400|}}
+Then we can use this configurations to create task.
+== Dataset ==
+For the dataset, you firstly need to change your dataset as rosbag file.
+Here we give you a new dataset folder structure:
+slam_hive_datasets
+For example: MH_01_easy sequence of EuRoC dataset, you should create a folder named "MH_01_easy", and then  in this folder, you should have following files:
+MH_01_easy.bag (need to be same name as folder name)
+groundtruth.txt: the groundtruth of this sequence (in TUM format)
+rosbag_play.py: used to play the rosbag file and remap some topic (this script is called by algorithm mapping.py script)
+------
+And as the topic name is different between algorithm required and dataset provided. So you need to set "algorithm remap" and "dataset remap" parameters to unify them.
+For example, the image topic name in dataset is "/cam0/image_raw", and in algorithm is "/camera/image_raw", the we need to set two paramters:
+{{:ndt4.png?1400|}}