Cleanlab/object-detection-tutorial

Object Detection Tutorial Dataset

Dataset Description

This dataset contains object detection annotations and predictions used in the cleanlab tutorial: Object Detection.

The dataset demonstrates how to use cleanlab to identify and correct label issues in object detection datasets, where labels consist of bounding boxes around objects in images.

Dataset Summary

• Total Examples: 118 images with bounding box annotations
• Task: Object detection with bounding boxes
• Files:
  • labels.pkl: Ground truth bounding box labels
  • predictions.pkl: Model predictions for bounding boxes
  • example_images.zip: Sample images for object detection

Dataset Structure

from huggingface_hub import hf_hub_download
import pickle
import zipfile

# Download labels
labels_path = hf_hub_download('Cleanlab/object-detection-tutorial', 'labels.pkl')
with open(labels_path, 'rb') as f:
    labels = pickle.load(f)

# Download predictions
predictions_path = hf_hub_download('Cleanlab/object-detection-tutorial', 'predictions.pkl')
with open(predictions_path, 'rb') as f:
    predictions = pickle.load(f)

# Download and extract images
images_path = hf_hub_download('Cleanlab/object-detection-tutorial', 'example_images.zip')
with zipfile.ZipFile(images_path, 'r') as zip_ref:
    zip_ref.extractall('example_images/')

Data Format

• labels.pkl: Dictionary containing ground truth bounding boxes in format [x_min, y_min, x_max, y_max, class_id]
• predictions.pkl: Dictionary containing predicted bounding boxes with confidence scores
• example_images.zip: Compressed folder containing image files

Dataset Creation

This dataset was created for educational purposes to demonstrate cleanlab's capabilities for detecting issues in object detection datasets, such as:

• Incorrectly labeled bounding boxes
• Missing annotations
• Poor quality predictions
• Annotation inconsistencies

Uses

Primary Use Case

This dataset is designed for:

1. Learning data-centric AI techniques for object detection
2. Demonstrating cleanlab's object detection issue detection
3. Teaching proper annotation quality assessment workflows

Example Usage

from huggingface_hub import hf_hub_download
import pickle
from cleanlab.object_detection.summary import object_detection_health_summary

# Download files
labels_path = hf_hub_download('Cleanlab/object-detection-tutorial', 'labels.pkl')
predictions_path = hf_hub_download('Cleanlab/object-detection-tutorial', 'predictions.pkl')

# Load data
with open(labels_path, 'rb') as f:
    labels = pickle.load(f)
with open(predictions_path, 'rb') as f:
    predictions = pickle.load(f)

# Use cleanlab to analyze object detection data quality
summary = object_detection_health_summary(labels, predictions)
print(summary)

Tutorial

For a complete tutorial using this dataset, see: Object Detection Tutorial

Licensing Information

MIT License

Citation

If you use this dataset in your research, please cite the cleanlab library:

@software{cleanlab,
  author = {Northcutt, Curtis G. and Athalye, Anish and Mueller, Jonas},
  title = {cleanlab},
  year = {2021},
  url = {https://github.com/cleanlab/cleanlab},
}

Contact

• Maintainers: Cleanlab Team
• Repository: https://github.com/cleanlab/cleanlab
• Documentation: https://docs.cleanlab.ai
• Issues: https://github.com/cleanlab/cleanlab/issues