Skip to content

Latest commit

 

History

History
 
 

VOC2007

README.md

Data preparation for Nucleus Detection module's training

Usage

Raw image annotation (not necessary if only using the provided annotations)

  1. Move the to-be-labelled images into JPEGImages/
  2. BatchRename.py – to rename JPEG images in JPEGImages/ to VOC2007 format
  3. Store the image IDs into ./ImageSets/Main/train.txt
  4. Using labelImg to create XML-format annotations into Annotations/
  5. voc_annotation.py – to convert the annotations from XML format to 2007_train.txt

Nucleus Detection module's training preparation

  1. MaskGenerator.py – to generate ground truth masks into ./Masks/ using the annotations in 2007_train.txt

Other files

  • PatchGenerator.py – to generate nucleus patches (only useful for visual inspection)
    • ground truth: using the annotations in 2007_train.txt, into ./Patches/GroundTruth/
    • prediction: using the results in ../FCRN/Results/Results_*.csv
  • Z_evalimages.py – to generate images into ./Patches/Z_eval/ for experts to label best focus
  • Annotations/ - contains the annotations of nucleus positions in XML format. The annotations are created using labelImg.
  • *.txt – contains the annotations of nucleus positions in txt format, which can be used for ../NucleusDetection/train.py