This is the ImageJ plugin for the density estimation by fully convolutional networks (DEFCoN) algorithm, an image processing tool for fluorescence spot counting. With this plugin, you can use trained DEFCoN models directly on images from within ImageJ.
- Download an already trained density map network model from the DEFCoN-ImageJ wiki. Unzip its contents to any directory you wish.
- Make a backup of your Fiji folder. (This is always a good idea before adding an update site.)
- Open Fiji and navigate to Help > Update.... Install any updates and restart Fiji if necessary.
- In the ImageJ Update dialog, click the Manage update sites button, scroll to the bottom of the list, and add http://sites.imagej.net/Kmdouglass under the URL column. You may give it any name you want, such as LEB-EPFL.
- Install all the updates and restart Fiji.
- Open an image or stack of images that contains fluorescent spots.
- Navigate to Plugins > DEFCoN and select Density map....
- Specify the location of the folder containing the DEFCoN density map network model that you downloaded and unzipped in the first step. This will be a folder (not a file!) that contains a TensorFlow SavedModelBundle and is generated by training the DEFCoN network.
- Click OK to compute a density map estimate from the images.
For more detailed instructions, please see the documentation.
http://defcon-imagej.readthedocs.io/en/latest/
Technical information about the design of DEFCoN may be found in the following preprint:
"Design Principles for Autonomous Illumination Control in Localization Microscopy," Marcel Štefko, Baptiste Ottino, Kyle M Douglass, Suliana Manley, bioRxiv 295519; doi: https://doi.org/10.1101/295519