Authors
Fidel A. G. Peña – Federal University of Pernambuco, Recife, Brazil
Pedro D. M. Fernandez – Federal University of Pernambuco, Recife, Brazil
Paul Tarr – California Institute of Technology, Pasadena, CA, USA
Ing R. Tsang – Federal University of Pernambuco, Recife, Brazil
Elliot Meyerowitz – California Institute of Technology, Pasadena, CA, USA
Alexandre Cunha – California Institute of Technology, Pasadena, CA, USA
Contact
Alexandre Cunha (cunha@caltech.edu)
Fidel A. G. Pena (fagp@cin.ufpe.br)
Algorithms
Additional information
We have trained our deep learning models exclusively on data provided by the challenge organizers, benefiting from the newly created silver truth. Details about the method we used in the competition can be found in our ISBI 2020 publication and its arXiv preprint. Our method is the latest in our developments of deep learning algorithms for cell segmentation. It is an improvement over our previous works using conceptually similar approaches on loss formulation, multiclass segmentation, data imbalance handling, and data augmentation, but with more sophisticated solutions. We are planning to experiment with other challenge datasets in the near future.