The purpose of this model is to classify a single document page image to define if it is the beginning page of the document or the middle/end page of a document. Single-page documents are classified as beginning page. It is a first step of the more general document boundary classification problem.
To generate the embeddings use google/siglip2-so400m-patch16-512 with no fine tuning.
You have a tiny script in generate_embeddings.py to generate a pickle file with the embeddings, provided a Pandas DataFrame tasks_df with a col "image_path" that contains all the images paths.
Then you can use the resulting embeddings with the model here uploaded.
Output meaning:
0 -> Middle or end page of a document
1 -> Beginning page of a document (or single-page document)