A Project for benchmarking existing SLAM algorithms on planes dataset. Benchmarking is done on an example of two ICL NUIM datasets: office and living room with respect to given ground truth trajectories. An aim of the project is to evaluate ate and rpe errors on both pre-annotated and frontend processed datasets, find key differences in the results and give minimum requirements to the frontend algorithm.
Benchmarks a trajectory, built by an algorithm
positional arguments:
main_data Directory where main information files are stored
annot Directory where color images are stored
{1,2,3} living room = 1, office = 2, point clouds = 3
first_node From what node algorithm should start
first_gt_node From what node gt references start
num_of_nodes Number of needed nodes
ds_filename_gt Filename of a file with gt references