Home

NOTE: in ubuntu run everything with sudo

SLAMBench is a framework that supports a wide range of sensors and datasets, allowing the evaluation of SLAM systems using qualitative and quantitative metrics, with plug-and-play algorithm support. The current iteration, SLAMBench4.0 adds lidar support and introduces container approach for SLAMBench, and contribute container implementations of the previously supported SLAM systems. For more details on Lidar support see this page. For more details on the benefits of containers please check out this page

SLAMBench can be either used with Docker containers or natively on the host device. For Container installation please see the instructions here. An example of using the container version of SLAMBench is here. NOTE: algorithms utilizing cuda do not currently have container versions

To use SLAMBench without docker have a look here

Compared with other benchmark tools, SLAMBench offers full control over the algorithm’s execution flow. It precisely calculates metrics for each frame and detects frames with significant relative pose errors (RPE). Furthermore, SLAMBench offers the capability to restart the algorithm from these pinpointed frames and incorporates an interface for the extraction and visualization of internal algorithmic processes. The last section on this page contains instructions on visualisation.