Can Self-Supervised Representation Learning Methods Withstand Distribution Shifts and Corruptions?
Prakash Chandra ChhipaJohan Rodahl HolmgrenKanjar DeRajkumar SainiMarcus LiwickiPublished in: ICCV (Workshops) (2023)
Keyphrases
- significant improvement
- empirical studies
- feature construction
- learning process
- prior knowledge
- knowledge acquisition
- learning algorithm
- learning models
- neural nets
- learning tasks
- benchmark datasets
- supervised learning
- training set
- preprocessing
- object recognition
- qualitative and quantitative
- inductive learning
- face recognition
- e learning
- data sets
- sample selection bias