Alam2019 - Machine learning approach of automatic identification and counting of blood cells

This model is used for automatic identification and counting of three types of blood cells: Red Blood Cells (RBC), White Blood Cells (WBC) and Platelet (Platelets) using the ‘you only look once’ (YOLO) object detection and classification algorithm with some additions to remove overannotation. The YOLO framework has been trained with a modified configuration BCCD Dataset of blood smear images to automatically identify and count red blood cells, white blood cells, and platelets.
Postprocessing with k-nearest neighbor (KNN) and intersection over union (IOU) approach reduces issues with multiple annotation of platelets.
The original code was extended to save the trained YoloV2 network state into the protobuf format. This is then used to generate the ONNX model, containing the weigths. Additional code was added to implement the inference step for image annotation based on the ONNX model, as well as the post-processing logic as used on the original model output. Dependencies have been documented explicitly using a conda environment.yml file to simplify reproducibility.

Original GitHub repository: https://github.com/MahmudulAlam/Automatic-Identification-and-Counting-of-Blood-Cells
GitHub repository: https://github.com/nilshoffmann/Automatic-Identification-and-Counting-of-Blood-Cells