Stable classification of diabetic structures from incorrectly labeled OCTA en face images using multi instance learning.
Philipp MattenJulius SchererThomas SchleglJonas NienhausHeiko StinoAndreas PollreiszWolfgang DrexlerRainer A. LeitgebTilman SchmollPublished in: Medical Imaging: Image Processing (2023)
Keyphrases
- face images
- multi instance learning
- training set
- face recognition
- feature vectors
- feature extraction
- multi instance
- training samples
- human faces
- classification accuracy
- principal component analysis
- gender recognition
- facial expressions
- facial features
- decision trees
- training data
- image classification
- supervised learning
- support vector
- input image
- feature space
- feature selection
- machine learning
- face verification
- face databases
- class labels
- benchmark datasets
- support vector machine
- feature set
- test images
- semi supervised learning
- pattern recognition
- high resolution
- active learning
- model selection