We provide the train and test splits of our newly curated BLC28 dataset to facilitate efficient evaluation and inspection during the review process.
BLC28 comprises 16,600 images spanning 28 blood cell categories (13 RBCs, 12 WBCs, and 3 Platelet classes), forming a balanced dataset.
The samples were selected to capture intra-class variability, including differences in:
cell morphology and structural patterns
staining intensity
imaging conditions and illumination
cell orientation and scale… See the full description on the dataset page:
https://huggingface.co/datasets/blcdata/BloodCellBank-Atlas.
