4DContrast: Contrastive Learning with Dynamic Correspondences for 3D Scene Understanding
This work addresses 3D scene understanding for applications like robotics or autonomous systems, presenting an incremental improvement by incorporating dynamic cues into existing contrastive learning frameworks.
The paper tackles the problem of learning 3D representations for scene understanding by using unsupervised pre-training with dynamic object priors, resulting in improved performance in downstream tasks like semantic segmentation and object detection, especially in data-scarce scenarios.
We present a new approach to instill 4D dynamic object priors into learned 3D representations by unsupervised pre-training. We observe that dynamic movement of an object through an environment provides important cues about its objectness, and thus propose to imbue learned 3D representations with such dynamic understanding, that can then be effectively transferred to improved performance in downstream 3D semantic scene understanding tasks. We propose a new data augmentation scheme leveraging synthetic 3D shapes moving in static 3D environments, and employ contrastive learning under 3D-4D constraints that encode 4D invariances into the learned 3D representations. Experiments demonstrate that our unsupervised representation learning results in improvement in downstream 3D semantic segmentation, object detection, and instance segmentation tasks, and moreover, notably improves performance in data-scarce scenarios.