Kyuwon Kim

Anil Armagan, Albert Saà-Garriga, Bruno Manganelli et al.

Gaussian Splatting (GS) is a popular approach for 3D reconstruction, mostly due to its ability to converge reasonably fast, faithfully represent the scene and render (novel) views in a fast fashion. However, it suffers from large storage and memory requirements, and its training speed still lags behind the hash-grid based radiance field approaches (e.g. Instant-NGP), which makes it especially difficult to deploy them in robotics scenarios, where 3D reconstruction is crucial for accurate operation. In this paper, we propose GSta that dynamically identifies Gaussians that have converged well during training, based on their positional and color gradient norms. By forcing such Gaussians into a siesta and stopping their updates (freezing) during training, we improve training speed with competitive accuracy compared to state of the art. We also propose an early stopping mechanism based on the PSNR values computed on a subset of training images. Combined with other improvements, such as integrating a learning rate scheduler, GSta achieves an improved Pareto front in convergence speed, memory and storage requirements, while preserving quality. We also show that GSta can improve other methods and complement orthogonal approaches in efficiency improvement; once combined with Trick-GS, GSta achieves up to 5x faster training, 16x smaller disk size compared to vanilla GS, while having comparable accuracy and consuming only half the peak memory. More visualisations are available at https://anilarmagan.github.io/SRUK-GSta.

Kyuwon Kim, Junhyuck Im, Gyuin Jee

Vehicle localization in highway tunnels is a challenging issue for autonomous vehicle navigation. Since GPS signals from satellites cannot be received inside a highway tunnel, map-aided localization is essential. However, the environment around the tunnel is composed mostly of an elliptical wall. Thereby, the unique feature points for map matching are few unlike the case outdoors. As a result, it is a very difficult condition to perform vehicle navigation in the tunnel with existing map-aided localization. In this paper, we propose tunnel facility-based precise vehicle localization in highway tunnels using 3D LIDAR. For vehicle localization in a highway tunnel, a point landmark map that stores the center points of tunnel facilities and a probability distribution map that stores the probability distributions of the lane markings are used. Point landmark-based localization is possible regardless of the number of feature points, if only representative points of an object can be extracted. Therefore, it is a suitable localization method for highway tunnels where the feature points are few. The tunnel facility points were extracted using 3D LIDAR. Position estimation is conducted using an EKF-based navigation filter. The proposed localization algorithm is verified through experiments using actual highway driving data. The experimental results verify that the tunnel facility-based vehicle localization yields precise results in real time.