3D-JEPA: A Joint Embedding Predictive Architecture for 3D Self-Supervised Representation Learning
This work addresses the problem of biased or inefficient 3D representation learning for computer vision tasks, offering a novel method that improves accuracy and efficiency, though it is incremental relative to prior joint embedding approaches.
The paper tackles the limitations of existing 3D self-supervised representation learning methods by introducing 3D-JEPA, a non-generative framework that uses a multi-block sampling strategy and context-aware decoder to predict target block representations from context blocks, achieving 88.65% accuracy on PB_T50_RS with 150 pretraining epochs.
Invariance-based and generative methods have shown a conspicuous performance for 3D self-supervised representation learning (SSRL). However, the former relies on hand-crafted data augmentations that introduce bias not universally applicable to all downstream tasks, and the latter indiscriminately reconstructs masked regions, resulting in irrelevant details being saved in the representation space. To solve the problem above, we introduce 3D-JEPA, a novel non-generative 3D SSRL framework. Specifically, we propose a multi-block sampling strategy that produces a sufficiently informative context block and several representative target blocks. We present the context-aware decoder to enhance the reconstruction of the target blocks. Concretely, the context information is fed to the decoder continuously, facilitating the encoder in learning semantic modeling rather than memorizing the context information related to target blocks. Overall, 3D-JEPA predicts the representation of target blocks from a context block using the encoder and context-aware decoder architecture. Various downstream tasks on different datasets demonstrate 3D-JEPA's effectiveness and efficiency, achieving higher accuracy with fewer pretraining epochs, e.g., 88.65% accuracy on PB_T50_RS with 150 pretraining epochs.