Compared with previous 3D reconstruction methods like Nerf, recent Generalizable 3D Gaussian Splatting (G-3DGS) methods demonstrate impressive efficiency even in the sparse-view setting. However, the promising reconstruction performance of existing G-3DGS methods relies heavily on accurate multi-view feature matching, which is quite challenging. Especially for the scenes that have many non-overlapping areas between various views and contain numerous similar regions, the matching performance of existing methods is poor and the reconstruction precision is limited. To address this problem, we develop a strategy that utilizes a predicted depth confidence map to guide accurate local feature matching. In addition, we propose to utilize the knowledge of existing monocular depth estimation models as prior to boost the depth estimation precision in non-overlapping areas between views. Combining the proposed strategies, we present a novel G-3DGS method named TranSplat, which obtains the best performance on both the RealEstate10K and ACID benchmarks while maintaining competitive speed and presenting strong cross-dataset generalization ability.

One of the most challenging problems in federated learning is the convergence speed problem caused by heterogeneity. We propose a novel structure called FedSD, a new method to accelerate the model convergence. We change the one-stage-cycle iteration structure to a 2-stage-cycle one to get the latest global gradient descent direction which can guide the model training direction. We instantiate algorithms using FedSD to improve the performance of experiments on several public datasets. Our empirical studies validate the excellent performance of FedSD.