BCData: A Large-Scale Dataset and Benchmark for Cell Detection and Counting

• Here is the website to release the BCData dataset for our work, BCData: A Large-Scale Dataset and Benchmark for Cell Detection and Counting (MICCAI 2020).

• Please click here to download the BCData dataset for cell detection and counting.

• The initial images are in the BCData/images folder, while the corresponding annotations are in the BCData/annotations folder. The dataset is separated into three parts: train, validataion and test. The annotations of the positive and negative tumor cells are in the positive subfolder and negative subfolder, respectively.

• Examples:

1) For the image with the path of BCData/images/train/10.png in the train set, the path for its annotations for the positive tumor cells is BCData/annotations/train/positive/10.h5, and the path for its annotations for the negative tumor cells is BCData/annotations/train/negative/10.h5.

2) For the image with the path of BCData/images/validation/10.png in the validation set, the path for its annotations for the positive tumor cells is BCData/annotations/validation/positive/10.h5, and the path for its annotations for the negative tumor cells is BCData/annotations/validation/negative/10.h5.

• Example script for loading the annotation (in Python):

import h5py

import numpy as np

gt_path = "Path of the corresponding .h5 file"

gt_file = h5py.File(gt_path)

coordinates = np.asarray(gt_file['coordinates'])