COVID-19: Face Mask Detection
Detect people wearing face masks in an image/video.
Challenge
About
The deadly COVID-19 is spreading exponentially and is causing havoc in our lives. Social distancing and wearing face masks are the only effective strategy against it. You as data scientists have a chance to play a key role in this battle. Computer Vision applications to detect whether a person is wearing a mask are of utmost importance.
The Challenge
You are required to build an AI model that detects people in an image and/or live video and classify whether they are wearing a face mask or not.
The task is to devise a face mask classifier, divided into two parts:
- Face detection
- Masked Classification
Dataset
The dataset consists of 5800+ RGB images, with two classes namely “face” and “face_mask”. The classification works on the faces present in an image. Training and Validation XML files provide annotations for the images giving details about location of faces and their classification based on presence of masks.
The directory structure of files is as follows:
.
└── COVID19-Classification
├── Training
│ ├── images
│ │ ├── img1.jpg
│ │ ├── img2.jpg
│ │ └── ...
│ └── annotations
│ └── training-images.xml
├── Validation
│ ├── images
│ │ ├── img1.jpg
│ │ ├── img2.jpg
│ │ └── ...
│ └── annotations
│ └── validation-images.xml
├── README.txt
└── Example
├── sample-image.jpg
└── sample-result.xml
│ └── ...
Submission
The output of the model should be box annotations for face co-ordinates in an image and their associated class. “Example” directory provides sample results to study.
Details regarding the submission are provided in README.txt
The final submission should include:
- Model Files (in formats supported by the above frameworks)
- requirements.txt (providing details of modules required to run your submission)
- Code Execution file (including training and testing ) (.py, .ipynb, ...,etc.)
Other files may be included for purposes of code modularity.
NOTE:
Currently only Python-framework based models are accepted:
- Tensorflow
- Pytorch
- Keras
The Judgement
The results will be declared by 5th of June 2020.
The criteria of judgement will be:
- Accuracy of detection algorithm
- Accuracy of classification algorithm
- Speed - Accuracy trade off
Rules
- Submission must not include copyrighted code. If violation is found, submission will be rejected.
- The submission should be in a proper format as described by README.txt (included in the dataset).
- Late submission will not be accepted beyond provided deadline (Indian Standard Time).
FAQs
Q. How do I apply for this Hackathon?
A. Login to your Dockship account, and click on "Submit Proposal" button on the current page.
Q. I've submitted the proposal, but received no response.
A.. Your proposal may be under review or the organisation has decided not to approve your proposal.
Q. My proposal is accepted, how do I download the dataset?
A.. After receiving the acceptance email, goto your "Dashboard" and select "My Proposals" from the sidebar. Select the correct Hackathon and click on "Download Dataset".
Q. Can I make multiple submissions?
A. Yes, you can submit your solutions multiple times.
Q. Where will the results be declared?
A. You will receive an email regarding the results on the mentioned result date.
Q. How will the prize money be awarded?
A. The prize money will be credited to your bank account within 7 days of declaration of results. You will have to complete your KYC before the amount can be credited.
Q. Can we apply as a team?
A. Yes you can. However, only the one team member can apply from a team.
Q. I've other queries, where can I get support?
A. Visit our support page: https://dockship.freshdesk.com/support/home for more FAQs or raise a support ticket.
Citations
Dataset Source:
- Real -World-Masked-Face-Dataset (https://github.com/X-zhangyang/Real-World-Masked-Face-Dataset)
- maskdetection (https://github.com/didi/maskdetection)
Challenge Started
Challenge Ended
