xjw00654/Micro-Nucleus-Detection

YOLO-v3+EfficientNet-b2

Micro-Nucleus Detection

Method

1. Using the YOLO-v3 for detection of all cells.
2. Using the EfficientNet-b2 for classification of cell types .

p.s. Using this separated 2-step detection rather than the end-2-end RCNN workflow for better finetuning model and modifying dataset.

File Tree

cls_m[0-5]_* is files for the classification workflow step 0 to 5.
train.py is the YOLO train code.
detect.py is the YOLO single image visualization code.
test.py is the YOLO performance evaluation code.
[train|validation]_list.txt is the train|validation data list.

Result