364 microscopic images of blood smears containing labeled bounding boxes for three distinct cell types: Red Blood Cells (RBC), White Blood Cells (WBC), and Platelets. The dataset utilizes the Pascal VOC XML format to provide spatial coordinates and class labels for each cell instance found in the images.
Use Cases
- Train an object detection model to localize blood cells using the 'bndbox' coordinates and 'name' tags in the XML files.
- Develop automated hematology counting tools by aggregating the number of 'object' entries per image for each cell class.
- Benchmark image preprocessing techniques on microscopic data to improve detection accuracy of smaller 'Platelet' instances.
Strengths
- 364 microscopic images of blood smears in JPEG format.
- Bounding box annotations for three classes: RBC, WBC, and Platelets.
- Annotations provided in Pascal VOC XML format including 'xmin', 'ymin', 'xmax', and 'ymax' coordinates.
- Includes a total of 4,888 individual cell instances across the image set.