Public Mask Placement Computer Vision Project
Updated 2 years ago
Metrics
Intro
This Computer Vision repository contains images from several sources:
- 346 original images from https://universe.roboflow.com/features-dataset/incorrect-mask-outside-mosiac-clean
- 2439 original images from https://universe.roboflow.com/ditworkspace/face-mask-jk4nr
I've applied dataset augmentation with the following parameters, to better respond to different types of video feeds / security cameras and such:
- Outputs per training example: 5
- Grayscale: Apply to 10% of images
- Blur: Up to 1.25px
- Bounding Box: Noise: Up to 1% of pixels
I attempt to also have the model ready for B/W and grayscale security feeds, blur in case there are faces out of focus when trying to run inference in real time (not very high, only 1.25px), and 1% noise to adapt to live streams with a lower-than-average bitrate.
I've generated 5x images than the original dataset. In total, augmenting the data, I produced:
- 10.000+ images in the training dataset (94%)
- 478 images for validation (4%)
- 216 images for testing (2%)
I aim to expand this dataset with YouTube videos in the future and add these samples into the testing set, to show how well the model performs with real-life examples.
Use This Trained Model
Try it in your browser, or deploy via our Hosted Inference API and other deployment methods.
Build Computer Vision Applications Faster with Supervision
Visualize and process your model results with our reusable computer vision tools.
Cite This Project
If you use this dataset in a research paper, please cite it using the following BibTeX:
@misc{
public-mask-placement_dataset,
title = { Public Mask Placement Dataset },
type = { Open Source Dataset },
author = { jasperan },
howpublished = { \url{ https://universe.roboflow.com/jasperan/public-mask-placement } },
url = { https://universe.roboflow.com/jasperan/public-mask-placement },
journal = { Roboflow Universe },
publisher = { Roboflow },
year = { 2023 },
month = { feb },
note = { visited on 2024-12-22 },
}