Skip to content

Diagrams Transformers

Wassim edited this page Jul 9, 2023 · 3 revisions

Images Transformers

  • CNN models : VGG, ResNet, Inception
  • OpenCV : SIFT, SURF, ORB
  • FastText
  • ImageNet
  • Hugging face transformers (pre-trained models) : ViT, DeiT

Diagrams

  • A Diagram can be interpreted from an image with text identification and OCR, but that is the least reliable approach
  • A Diagram can be parsed from an SVG file that already contains clear text and shapes, still might require a significant pre-processing
  • A Diagram can be parsed from an XML file of a dedicated tool, e.g. Drawio (diagrams.net) that already contains meaningful information
  • A reasonable intermediate step is to transform the Diagram to a graph and then use known graph Embeddings
    • Node2Vec
    • Graph2Vec
    • GraphSAGE

References

Clone this wiki locally