HW3 Introduction: Nuclei segmentation
The proposed challenge is a nuclei segmentation, which contains two parts:
- Transform the original train data format to coco dataset format for detectron2 training Segment the nuclei of bounding boxes
- Nuclear segmentation dataset contains 24 training images with 14,598 nuclear and 6 test images with 2,360 nuclear Train an instance segmentation model to detect and segment all the nuclei in the image
This project uses the detectron2 pre-trained model to fix this challenge.
model : Mask-RCNN on detectron2
reference code from detectron2 official github