Detecting Cells in Histopathology Images with a ResNet Ensemble Model

This paper presents a candidate solution that was submitted to the test phase of the OCELOT2023 challenge. This solution is based on an ensemble of trained fully convolutional ResNet-50 models that were developed using only the small field-of-view (FoV) images with cell-level annotations of the challenge dataset (the large FoV images with tissue-level annotations were not used). The submitted model achieved a F1-score of 0.673 on the evaluation set of the validation phase. The code to run our submitted trained model is available at: https://github.com/CTPLab/OCELOT2023_P4ResNet50 .