---
dataset_info:
  features:
  - name: image_id
    dtype: string
  - name: image
    dtype: image
  - name: annotations
    struct:
    - name: image
      dtype: string
    - name: prefix
      dtype: string
    - name: suffix
      dtype: string
  splits:
  - name: train
    num_bytes: 5506310
    num_examples: 255
  - name: test
    num_bytes: 772572
    num_examples: 36
  - name: validation
    num_bytes: 1570520
    num_examples: 73
  download_size: 7729403
  dataset_size: 7849402
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: test
    path: data/test-*
  - split: validation
    path: data/validation-*
license: cc-by-4.0
---


# Dataset Card for roboflow-bccd-paligemma

## Dataset Description

- **Repository:** https://huggingface.co/datasets/dwb2023/roboflow-bccd-paligemma
- **Paper:** N/A

### Dataset Summary

This dataset, roboflow-bccd-paligemma, is a modified version of the BCCD (Blood Cell Count and Detection) dataset. It contains blood cell images annotated for object detection tasks, specifically targeting three types of blood cells:

1. Platelets
2. Red Blood Cells (RBC)
3. White Blood Cells (WBC)

Key features of the dataset:
- Total of 364 annotated images across train, validation, and test splits
- Bounding box annotations for each detected cell
- Labels identifying the cell types

The dataset is structured to support object detection tasks in the medical imaging domain, particularly for blood cell analysis.

It's crucial to note that this dataset is a derivative work based on the original BCCD dataset. When using this dataset, proper attribution is essential. Please use the citation provided at the end of this card in any work that utilizes this data.

### Supported Tasks and Leaderboards

- **Tasks:** Object Detection

### Languages

The dataset uses English labels.

## Dataset Structure

### Data Instances

A typical data instance contains:
- An image of blood cells
- Bounding box annotations for detected cells
- Labels identifying the cell types (Platelets, RBC, WBC)

### Data Fields

- `image_id`: Unique identifier for each image
- `image`: The blood cell image
- `annotations`: Contains annotation details

The `annotations` field has the following structure:

| Field Name | Description | Data Type |
|------------|-------------|-----------|
| image      | Identifier for the image being annotated | string |
| prefix     | Standard prefix for all annotations, typically "detect Platelets ; RBC ; WBC" | string |
| suffix     | Contains the actual annotation data, including bounding box coordinates and cell type labels | string |

The `suffix` field contains multiple annotations for each image, separated by semicolons. Each annotation typically follows this format:

`<locx1><locy1><locx2><locy2> CellType`

Where:
- `<locx1><locy1><locx2><locy2>` represent the bounding box coordinates
- `CellType` is one of: RBC, WBC, or Platelets

Example of a complete annotation:
```
prefix: "detect Platelets ; RBC ; WBC"
suffix: "<loc0834><loc0158><loc1004><loc0347> RBC ; <loc0586><loc0193><loc0827><loc0403> RBC ; <loc0040><loc0169><loc0290><loc0377> WBC"
```

### Data Splits

The dataset is divided into three splits:
- Train: 255 images
- Validation: 73 images  
- Test: 36 images

## Considerations for Using the Data

### Social Impact of Dataset

This dataset could potentially aid in automating blood cell counting and classification, which may improve efficiency in medical diagnostics. However, as with any medical-related AI application, care must be taken to ensure accuracy and proper validation before clinical use.

## Additional Information

### Licensing Information

This dataset is licensed under Apache 2.0.

### Citation Information

If you use this dataset in your research, please cite it as:

```
@misc{
                            bccd-ouzjz_dataset,
                            title = { bccd Dataset },
                            type = { Open Source Dataset },
                            author = { Roboflow 100 },
                            howpublished = { \url{ https://universe.roboflow.com/roboflow-100/bccd-ouzjz } },
                            url = { https://universe.roboflow.com/roboflow-100/bccd-ouzjz },
                            journal = { Roboflow Universe },
                            publisher = { Roboflow },
                            year = { 2023 },
                            month = { may },
                            note = { visited on 2024-08-02 },
                            }
```