Avian Blood Cell dataset

This dataset contains 4,600 avian blood smear microscopy images for automated blood cell analysis. It includes five cell types: RBCs, thrombocytes, lymphocytes, heterophils, and monocytes. The dataset is split into 1,000 labeled images and 3,600 unlabeled images. Of the labeled images, 500 contain both segmentation masks and bounding box annotations, while 500 contain bounding box annotations only. The original image resolutions range from 1280 × 720 to 3264 × 1840 pixels but for the experiments described in the associated paper, the 1000 annotated images were resized to 640 × 640 pixels. The dataset is suitable for object detection, instance segmentation, cell classification, representation learning, and benchmarking data-efficient methods for avian hematology.