Counting objects is a learning task common to several applications such as video surveillance, agriculture 4.0, life sciences and medicine. However, this task is often carried out manually by human operators, involving a significant effort in terms of time and human resources. At the same time, this procedure is typically prone to errors due to fatigue of the annotators.
In this seminar, we present a Deep Learning approach to automate recognizing and counting neuronal cells in microscopic fluorescence images. First, we introduce the Fluorescent Neuronal Cells open dataset and describe its peculiar traits and challenges. Then, we discuss neural network architecture tuned specifically for this use case, the cell-ResUnet (c-ResUnet).
Finally, we focus on performance assessment from different perspectives and briefly outline some possible future research lines.
For more details, please refer to:
Thesis: http://amsdottorato.unibo.it/10016/1/thesis_CLISSA_DSC.pdf
Article: https://www.nature.com/articles/s41598-021-01929-5.pdf
Dataset: http://amsacta.unibo.it/6706/
Speaker:
Luca Clissa(Istituto Nazionale di Fisica Nucleare)
