Inclusion Request for Customizable SLAM Robustness Benchmark (Gaussian Splatting vs NeRF representation for instant 3D reconstruction and localization)
Xiaohao-Xu opened this issue · comments
Hey guys,
We sincerely appreciate the value of your repository! We would like to kindly request the inclusion of our customizable SLAM robustness benchmark, which incorporates a noisy data simulation pipeline for a comprehensive evaluation of SLAM performance.
Our benchmark mainly focuses on evaluating the robustness of the latest monocular/multi-modal Neural-based SLAM models, including NeRF-based and Gaussian-Splatting-based SLAM, which reveals GS representation is more robust than NeRF representation under sensor corruptions for instant 3D reconstruction and localization.
We greatly appreciate your consideration. Thanks!
Hey,
thanks for the ping. Might be a valuable addition. I am a bit hesitant because I would say it is very strongly tied to SLAM in general and less specific to gaussian splatting.
I have thought about adding a data category. What do you think were this would fit best in?
Sure thing! I think it would be awesome to add a new category 'Data' or 'Benchmark.' It'll make things easier to find and organize.
Ok great, I will add it over the weekend!
I introduced a data section. But I think it is better to link it when it is published. Otherwise I don't see much of value here as it is also not very specific to 3DGS. Please ping me when you have uploaded the dataset.