toc: true title: Robust RegNet

tags: [‘temp’]


Robust RegNet

  • Vision Models are More Robust and Fair When Pretrained on Uncurated Images Without Supervision
  • Unsupervised Learning
  • Discriminative Self Supervised learning allows training models on any random group of internet images, and possibly recover salient information that helps differentiate between the images
  • ImageNet
  • object-centric Features that perform on par with supervised Features on most object-centric downstream tasks
  • learn any salient and more representative information present in diverse unbounded set of images from across the globe
  • without any data pre-processing or prior assumptions about what we want the model to learn
  • RegNet
  • scaled to a dense 10 billion parameters
  • pre-trained using the SwaV self-supervised method on a large collection of 1 billion randomly selected public images from Instagram with a diversity of gender, ethnicity, cultures, and locations
  • captures well semantic information
  • captures information about artistic style and learns salient information such as geo-locations and multilingual word embeddings based on visual content only.
  • large-scale self-supervised pre-training yields more robust, fair, less harmful, and less biased results than supervised models or models trained on object centric datasets such as ImageNet