Cheese2School is a face-recognition application for mobile devices that allows to count how many people there are in a room or an open space, by simply taking a picture of the group. The application was developed to help teachers in Burkina Faso record students’ attendance, but can be used more widely to instantly perform a head-count. If properly trained, the app can also distinguish between male and female attendees, but this requires the availability of a relatively big collection of images already classified.
In its deployment in Burkina Faso, the app verifies the students present per class, categorizing them per gender: each of the faces is reworked using a specific artificial neural network to recognize the male or female gender. The data collected in schools are aggregated and centralized inside a database accessible by educational institutions, both at central government and local level. The app uses algorithms of artificial neural networks, drawing on the rich open-source tools already available. .
What is Machine Learning and AI?
Data analysis made some huge leaps in the last years, allowing software not only to see, but also to understand what is looking at. We analyzed some data sets for Classification problems, making our model actually distinguish between entities, being it some lexical notions, as in Natural Language Processing techniques, or plain pictures containing subjects. There are several tools at our disposition for the task, from some simple Probability Models to some more advanced Convolutional Neural Networks. As usual one can only predict efficiently an event which is generated from some discovered strong patterns, otherwise we go back to guessing. Artificial intelligence should and will enhance all those social interventions already done by thousands of people in the most critical areas of the globe, and I believe actually it is the best field where AI should be deployed. Using learning networks in video games is fun. Selling the right burger to the right person is nice. But what about those situations where time is not an option, nor errors are allowed. Situations where life itself depends on fast and precise reaction of those who can help.
Why use Cheese2School
If you need to perform a head count and want to skip the tedious one-by-one counting process because:
- you have no time
- you want to avoid mistakes
- you need to repeat it often
You need to count people in one of these typical situations:
- students in a class
- participants to a community meeting
- members of an assembly
- people in a gathering
Data are automatically inserted into our data collection system, DEWCO, and can therefore be easily integrated into a wider data collection activity You are concerned about privacy. C2S is GDPR compliant in that it does not store or transmit any sensitive data. All processing is done on the mobile device and then deleted. Only aggregated information is stored and transmitted.
Characteristics of c2s
Considering the context with constraints on both devices power and poor connectivity, the model MobileNet, proposed by Google, and framework Open Source TensorFlow, released by Google Brain, have been chosen for this work to carry out recognition and classification, the network has been trained with a lot of data. It uses the model MobileNet already trained by the database ImageNet, composed of more than 14 millions of images.
Cheese2School App Core Features
- Easy Digitization: Digitize your questionnaires for collecting data on mobile devices through our simple point-and-click user interface.
- Skip Logic: Configure skip logic to guide data entry and ensure that only relevant questions are posed to the respondent.
- On/offline Data Collection: When online, collected data is transmitted instantly . Otherwise, the data is transmitted when connectivity is restored.
- Validation Constraints: Set constraints such as range-checks and unique fields to ensure only quality data is entered in your database.
- Longitudinal Data: Linked forms and two-way syncing help you configure forms for longitudinal data collection with initial and follow-up visits.
- Data Analysis (BI) plug-in: Once the data has been entered into the platform, it will be possible to extract it with a module that will interface with the main data analysis tools.
- API Integration: Forward data collected to third-party applications like DHIS2 and Tableau using a webhook push API.
- Security: Your data is automatically secured on both the mobile and web applications using industry standard security.
- GPS: Automatically tag records with GPS data for quality control and mapping. Visualize GPS points on Google Maps.
- Realtime Analysis: Analyze and review your data in real-time with frequency distributions, summary statistics, dashboards and geo-maps.