Approaching deep learning based object detection in microscopy images to non-expert users

Abstract

In this project, we have first carried out a study of the state of the art in object detection with Deep Learning, and then we have designed and implemented an approach that is oriented to be run in a cloud service by non-expert users. More specifically, due to its possible applications in microscopy image analysis, a web-based solution that uses the state-of-the-art RetinaNet model has been developed in the open-source ZeroCostDL4Mic environment. Moreover, our implementation uses the TensorFlow 2 object detection API, that allows different backbone networks, and it has been accepted as part of the official ZeroCostDL4Mic platform. Finally, the evaluation of the proposed solution has been performed in a public dataset and compares positively with alternative state-of-the-art approaches.