PBC labeling Extended (pbc_improve)

Dataset Name: PBC labeling Extended (pbc_improve) Description: This dataset is an enhanced version of the public peripheral blood cell dataset "PBC labeling". The original dataset contains 8 classes of target cells (Basophils, Eosinophils, Erythroblast, Ig, Monocytes, Platelets, lymphocytes, neutrophils) but lacks annotations for erythrocytes (red blood cells). To support research on obstacle‑avoidance path planning for optical tweezers‑based cell manipulation, we manually annotated erythrocytes using LabelImg and performed data cleaning to remove low‑quality or mislabeled samples. The final dataset, named PBC labeling Extended, comprises 9 cell classes (8 original target cells + erythrocytes). The dataset is split into training (1372 images), validation (356 images), and test (182 images) sets. All annotations follow the YOLO format: each image has a corresponding .txt file containing lines of class_id x_center y_center width height with normalized coordinates. This dataset is intended for training and evaluating deep learning‑based cell detection models, particularly for scenarios that require simultaneous recognition of multiple cell types in dense and partially overlapping microscopic images. Key Features: 9 cell classes (IDs 0–7: target cells; ID 8: erythrocyte) YOLO‑format annotations Balanced train/val/test splits Data cleaning applied to ensure quality License: CC BY 4.0