synthval.features_extraction#
Module for feature extraction using various models from HuggingFace, including Rad-Dino, DinoV2, and MambaVision.
This module defines abstract and concrete classes for feature extraction from images, leveraging different pre-trained models available through the HuggingFace library. The module supports extracting features from images using models such as Rad-Dino, DinoV2, and MambaVision, with each extractor providing methods for single and batch feature extraction.
Classes#
- FeatureExtractor(abc.ABC)
Abstract base class for defining a feature extractor interface.
- RadDinoFeatureExtractor(FeatureExtractor)
Concrete feature extractor using the HuggingFace Rad-Dino model.
- DinoV2FeatureExtractor(FeatureExtractor)
Concrete feature extractor using models from the HuggingFace DinoV2 family.
- MambaFeatureExtractor(FeatureExtractor)
Concrete feature extractor using models from the HuggingFace MambaVision family.
- InceptionExtractor(FeatureExtractor)
Concrete feature extractor using traditional Inception models from the timm library.
Classes#
Abstract base class representing a generic feature extractor. |
|
Feature extractor using the HuggingFace model microsoft/rad-dino for extracting features from images. |
|
Feature extractor using models from the HuggingFace DinoV2 family |
|
Feature extractor using models from the HuggingFace MambaVision family |
|
Feature extractor using traditional Inception models from the timm library (https://huggingface.co/docs/timm/index). |