DEVELOP.md

Develop

Installation

We used PyTorch 1.9.1 on Ubuntu 22.04 LTS with Anaconda Python 3.7.

1. [Optional but recommended] Create a new Conda environment.

   conda create --name Map3D python=3.7
   

   And activate the environment.

   conda activate Map3D
   

2. Clone the Map3D repo

3. Install the requirements

4. Install ANTs:

   cd Map3D/Map3D-pipeline/ANTS
   sh installANTs.sh
   cd ..
   

Run Map3D Registration pipeline locally without docker

• The entire pipeline is at the Map3D-pipeline folder
• Create an empty folder in the Map3D-pipeline folder and name it as "input_png". Put folders that contain 10X magnification PNG files into "input_png" folder. For guidance and instruction for input data format requirement and data arrangement, please refer to DATA.md.
• Run python scripts in Map3D-pipeline folder as following orders:
  1. Step1_superglue.py
  2. Step2_ApplySGToMiddle.py
  3. Step3_matrix_npytomat.py
  4. Step4_SuperGlue+ANTs.py
  5. Step5_BigRecon_moveAllslicesToMiddle.py
• Note that Step2_ApplySGToMiddle.py, Step4_SuperGlue+ANTs.py, and Step5_BigRecon_moveAllslicesToMiddle.py require an argument. Example can be found in "Map3D Registration Demo" section at README.md.
• The output will be stored at "output" folder under Map3D-pipeline directory.