Harness the power of convolutional neural networks, realized through cutting edge artificial intelligence engines to advance your image processing results to new levels.
Powered by Google’s TensorFlow and Keras, Dragonfly gives users the power to develop new neural networks, but also to train, reuse, and repurpose existing models for advanced applications that will revolutionized your workflows.
Because Dragonfly’s trained neural networks behave like image filters, they are easy to preview, fast to apply, and simple to share for reuse. Publish your solutions to Dragonfly’s Infinite Toolbox for others to use; likewise, download models shared by others and apply directly to your data.
Extensible Library of Neural Networks
Dragonfly’s Deep Learning solution is bundled with pre-built and pre-trained neural networks, implementing such powerful solutions as UNet, DenseNet, FusionNet and many others.
Novice users find it easy to apply Dragonfly’s powerful segmentation features on select reference slices and then use those results to train existing neural networks. Those neural networks can, in turn, segment the rest of the same experimental image stack and subsequent ones, thereby saving countless hours of demanding work.
Advanced users can take advantage of built-in tools for drafting new networks or edit the activation functions and other node behavior in existing models. Moreover, users can code directly in Python with the Keras API or import pre-existing Keras models for direct integration in Dragonfly.
Ceramic matrix composite, imaged by synchrotron microCT at the Advanced Light Source beam line 8.3.2 at the Lawrence Berkeley National Laboratory.
Image courtesy of Aly Badran (University of Colorado).
Drosophila neural tissue imaged by serial section TEM. Courtesy of ISBI Challenge 2012.
Reference: Cardona A, Saalfeld S, Preibisch S, Schmid B, Cheng A, Pulokas J, Tomancak P, Hartenstein V. 2010. An Integrated Micro- and Macroarchitectural Analysis of the Drosophila Brain by Computer-Assisted Serial Section Electron Microscopy. PLoS Biol 8(10).
C. elegans imaged by serial section FIB-SEM.
Image courtesy of Dr. Keana Scott (National Institute of Standards and Technology).