Midgut_oocyst_segmentation

Perform oocyst segmentation in mercurochrome stained mosquito midguts

This oocyst segmentation model also powers the webtool at http://got2findthemall.org/

Requirements to run locally:

(1) Build tools
     Visual Studio build tools if using Windows:
     https://visualstudio.microsoft.com/visual-cpp-build-tools/

     Xcode if using MacOS

     Linux: you should be able to figure this out if you use Linux

(2) Python and packages

Python >3.8

Python packages:
Notes: the listed versions are tested to work. You can use pip to install all the packages listed here,
or create a conda environment using conda_env.yml supplied by this repo
     torch=1.9.1
     torchvision=0.10.0
     torchaudio=0.9.1
     pandas=1.3.3
     pycocotools=2.0.2
     dataclasses=0.6
     typing=3.7.4.3
     opencv-python=4.5.3.56
     xlsxwriter=3.0.1
     detectron2=0.5

For the detectron2 package, you can git clone the repo and install using (must have Git installed):
python -m pip install ‘git+’ –user

Instructions:

(1) Clone the repository

(2) unzip the two model weight files and keep it in the “model” directory:
     <1> model_0002399.MG.pth
       (unzip from model_0002399.MG.zip.001 and model_0002399.MG.zip.002)
     <2> model_0006199.pth
       (unzip from model_0006199.zip.001 and model_0006199.zip.002)

(3) prepare your own jpeg images and place them in a folder, or use the “test_images” folder

(4) run oocyst segementation with the following command:
     python oocyst_segmentation.py –dir [path to your folder]

(5) Four result files with the same prefix will be generated for each image
      count_N_size.xlsx               oocyst count, area and coordiate of each oocyst, average area
      [prefix].oocyst.jpg                oocyst annotated on the original image
      [prefix].midgut.jpg                midgut annotated on the original image
      [prefix].midgut.MASK.jpg       A full-resolution black-whight MASK of the midgut identified

GitHub

https://github.com/duopeng/midgut_oocyst_segmentation