Automatic Annotation for 3D Object Detection using 2D visual foundation model.

We propose an Mixture-of-Experts based continual learning method for Open-Vocabulary Segmentation to handle sequentially collected datasets.