*Paper Accepted by ECCV 2024*

Haozhi Cao, Yuecong Xu, Jianfei Yang, Pengyu Yin, Xingyu Ji, Shenghai Yuan, Lihua Xie

[OpenAccess] [Poster][Code]

• [2024.09] Full code released! Check our github repo for more details. 

• [2024.09] Installation and data preparation details released.  The full release will be available soon!

• [2024.08] We are now refactoring our code. Stay tuned for the full release!

• [2024.08] Our paper has been accepted by ECCV 2024🎉!

Multi-modal test-time adaptation (MM-TTA) is proposed to adapt models to an unlabeled target domain by leveraging the complementary multi-modal inputs in an online manner. Previous MM-TTA methods for 3D segmentation rely on predictions of cross-modal information in each input frame, while they ignore the fact that predictions of geometric neighborhoods within consecutive frames are highly correlated, leading to unstable predictions across time. To fulfill this gap, we propose ReLiable Spatial-temporal Voxels (Latte), an MM-TTA method that leverages reliable cross-modal spatial-temporal correspondences for multi-modal 3D segmentation. Motivated by the fact that reliable predictions should be consistent with their spatial-temporal correspondences, Latte aggregates consecutive frames in a slide window manner and constructs Spatial-Temopral (ST) voxels to capture temporally local prediction consistency for each modality. After filtering out ST voxels with high ST entropy, Latte conducts cross-modal learning for each point and pixel by attending to those with reliable and consistent predictions among both spatial and temporal neighborhoods. Experimental results show that Latte achieves state-of-the-art performance on three different MM-TTA benchmarks compared to previous MM-TTA or TTA methods.