Skip to content

Latest commit

 

History

History
10 lines (10 loc) · 1.64 KB

TODO.md

File metadata and controls

10 lines (10 loc) · 1.64 KB
Raw

TODO

V2

  • [] Different models will be implemented for face detection. For example SCRFD models from insightface etc.
  • [] Multiple image and video processing for face detection, followed by saving output files (.JSON or .npy) for each of the processed images and videos will be implemented.
  • [] New face recognition models will be implemented. For example FaceNet, GhostFaceNets, AdaFace etc.
  • [] In addition to the current Arcface R100 recognition model, models trained with different backbone and different datasets will be implemented.
  • [] Multiple image option will be implemented for facial attribute analysis. Output will be availabe for download format (.csv or .pkl).
  • [] A new model will be implemented for face parsing which is high quality and segments only the face. The output mask will be aligned to the original frame and given in the same format. For example, Graphonomy etc.
  • [] A structure that receives video input for landmark extraction will be implemented. Here, a high quality model that uses tracking algorithms and minimises flicker will be implemented. For example, SPIGA etc.
  • [] In ControlNet operations, a structure that receives multiple image and video inputs and records the outputs frame-by-frame will be implemented.