Face Mask Detection Dataset

The Face Mask Detection dataset is designed for computer vision projects that identify whether people wear face masks correctly. It contains 853 images and 4,072 labeled faces, making it useful for developing and evaluating face mask detection models.

The dataset includes three detection classes: with mask, without mask, and mask worn incorrectly. In addition, the images come with bounding-box annotations that help computer vision models locate faces and classify their mask status.

Therefore, researchers and developers can use this dataset for object detection and image-based AI projects. The annotations follow the PASCAL VOC format, which also makes the dataset suitable for workflows built with popular computer vision frameworks.

Key Features

• 853 images for face mask detection tasks.

• 4,072 labeled faces with object-level annotations.

• Three detection classes: with mask, without mask, and mask worn incorrectly.

• Bounding-box annotations for locating faces within images.

• PASCAL VOC/XML annotation format for object detection workflows.

• Suitable for computer vision and deep learning research.

• Useful for training and testing face mask detection models.

Classes Included

The dataset focuses on three face mask categories:

With Mask

This class represents faces where the person is wearing a mask correctly.

Without Mask

This class contains faces where no face mask is present.

Mask Worn Incorrectly

This category identifies faces where a person wears a mask incorrectly, such as when the mask does not properly cover the required facial area.

The dataset contains more examples of masked faces than incorrectly worn masks, so researchers should consider class imbalance when training models.

Applications of the Face Mask Detection Dataset

Face Mask Detection

Develop computer vision systems that detect whether a person is wearing a mask.

Object Detection

Train object detection models to locate faces and identify their mask status using bounding-box annotations.

Computer Vision Research

Use the images for experiments involving image recognition, object detection, and visual classification.

Deep Learning Projects

The dataset can support deep learning workflows involving models such as Faster R-CNN, YOLO, and other object detection architectures. Existing projects have used the dataset with models such as Faster R-CNN and YOLO-based approaches.

AI Model Evaluation

Researchers can use the annotated images to compare different approaches for detecting masked, unmasked, and incorrectly masked faces.

Why Use the Face Mask Detection Dataset?

The dataset combines real-world face images with object-level annotations, which makes it useful for practical computer vision experiments. Moreover, its three-class structure allows models to distinguish between correctly masked, unmasked, and incorrectly masked faces.

The bounding-box annotations also provide location information instead of only image-level labels. As a result, developers can build object detection systems that identify both the face location and its mask category.

However, the dataset has class imbalance and many faces are relatively small. Researchers should consider these factors during preprocessing, training, and evaluation.

Who Can Use This Dataset?

The Face Mask Detection dataset can benefit:

• Computer vision researchers

• Machine learning developers

• AI students and learners

• Deep learning practitioners

• Object detection researchers

• Data scientists

• AI project developers

• Researchers working on image recognition and surveillance systems

Conclusion

The Face Mask Detection dataset provides 853 images and 4,072 labeled faces across three mask-related categories. With bounding-box annotations and PASCAL VOC formatting, it offers a useful resource for computer vision and object detection projects.

Researchers and developers can explore this dataset for face mask detection, AI model training, image analysis, and computer vision research.

The dataset is available on Kaggle