only obtain image features?

Is there any method we can extract features of images ? Like in SAM-v1 where we can only use vit to encode image into features and discard the other parts of the model ?

The issue does not identify a file, test, or entry point. Start by reviewing the repository's example notebooks and inference path for the image encoder, then determine how image features should be exposed and what output would demonstrate that the feature-only workflow is complete.

Written by the indexing model from the issue text.