4 ms·
The image is subdivided by a grid and the resulting patches are fed through a linear encoder to get the token embeddings.
by bonoboTP 3y ago
The image is subdivided by a grid and the resulting patches are fed through a linear encoder to get the token embeddings.