Make sure you have Python installed, then install Tensorflow on system, then clone this repo.

Image classification with transfer learning from Google's Inveption v3 already done with models and labels uploaded in Dropbox links for brightfield and fluorescent. Both folders must be unzipped and placed directly in the cellimaging folder as indicated below.

|
---- /cellimaging
|    |
|    |
|    ---- /logs-brightfield
|    |    trained_graph.pb
|    |    trained_labels.txt
|    |
|    ---- /logs-fluorescence
|    |    trained_graph.pb
|    |    trained_labels.txt
|    |
|    ---- /images
|    |    firstimagec1t1.tiff
|    |    firstimagec12t1.tiff
|    |    secondimagec1t2.tiff
|    |    secondimagec12t2.tiff
|         ...
|

Ensure that requirements are installed by running

pip install -r requirements.txt

Note that logs-brightfield and logs-fluoresecence must maintain the same folder name. The image folder you are analyzing is not restricted by naming.

python gui.py

Select image folder and click submit. Progress is documented in terminal.

gui.py: runs user input for image folder with GUI

• for each timepoint/location, must have two corresponding images (brightfield and fluoresecent)

• naming of files must end with c1t* for brightfield and c2t* for fluorescent, otherwise same

• note: folder of images MUST be within the 'cellimaging' folder directory

getCounts.txt: macro for ImageJ to get coordinates of selected points (for centers of droplets)

getcenters.py: automated cell selection and numbering with Canny edge detection and modified Hough Circle transform with thresholded area and circularity

blobdetector.py: example of image detection with no machine learning

boundaries-original.py: original code with keras model for wellplate (in progress)

boundaries.py: only with google inception v3 transfer learning

canny.py: detecting with Watershed and distance transform (in progress)

classify.py: identifies brightfield photos

classify2.py: identifies fluoresecent photos

convert.py: 90, 180, 270 rotation with reflection (vertical and horizontal) for initial manual labelling

helpers.py: required helper functions for boundaries