Figure 2. Overview of Plant-MAE. (IMAGE)
Caption
(a) The main structure of the self-supervised pretraining model, including three parts—a token embedding at the forefront, followed by a hierarchical encoder–decoder and a point reconstruction module. (b) The main structure of the fully supervised fine-tuning model, which includes a pretrained encoder and a segmentation head. (c) Demonstration of the encoder structure used in (a) and (b), which consists of four MAFEBs linked by residual connections.
Credit
The authors
Usage Restrictions
Credit must be given to the creator.
License
CC BY