How to extract semantic point features based on coordinates (x, y)

<img width="390" alt="image" src="https://github.com/user-attachments/assets/c24a607e-f9dd-4048-a05e-0699ff0e9720" />
In the DIFT repository, only the script for extracting features from images is provided. How is the extraction of semantic point features based on coordinates (x, y) implemented, as you mentioned?