{
  "id": "1911.09652",
  "version": "v1",
  "published": "2019-11-20T07:36:45.000Z",
  "updated": "2019-11-20T07:36:45.000Z",
  "title": "Unsupervised Domain Adaptation by Optical Flow Augmentation in Semantic Segmentation",
  "authors": [
    "Oluwafemi Azeez"
  ],
  "comment": "arXiv admin note: text overlap with arXiv:1910.10369 by other authors",
  "categories": [
    "cs.CV",
    "cs.LG",
    "eess.IV"
  ],
  "abstract": "It is expensive to generate real-life image labels and there is a domain gap between real-life and simulated images, hence a model trained on the latter cannot adapt to the former. Solving this can totally eliminate the need for labeling real-life datasets completely. Class balanced self-training is one of the existing techniques that attempt to reduce the domain gap. Moreover, augmenting RGB with flow maps has improved performance in simple semantic segmentation and geometry is preserved across domains. Hence, by augmenting images with dense optical flow map, domain adaptation in semantic segmentation can be improved.",
  "revisions": [
    {
      "version": "v1",
      "updated": "2019-11-20T07:36:45.000Z"
    }
  ],
  "analyses": {
    "keywords": [
      "unsupervised domain adaptation",
      "optical flow augmentation",
      "generate real-life image labels",
      "domain gap",
      "simple semantic segmentation"
    ],
    "note": {
      "typesetting": "TeX",
      "pages": 0,
      "language": "en",
      "license": "arXiv",
      "status": "editable"
    }
  }
}