Poor sanitation has a terrible impact not only on health but also on the education of children - particularly when females drop out. The numbers scream agonizingly – 10% of global diseases are due to lack of sanitation and the girls' dropout ratio is at a staggering 23%. Globally, the annual loss is estimated at ~260 billion USD, and closer home, it also adversely impacts the Swachh Bharat Mission of the Government of India.

It’s impossible to manage the numbers (toilets) through manual inspection. In this case, manual interventions are fraught with all kinds of challenges – human bias (politically motivated to make the states look good), error, and unsystematic monitoring, to mention a few. 

Can AI Solve This Problem?

Yes, says TCS.

And, they have a comprehensive automated Sanitation Inspection solution for grading sanitation components, which is also the first of its kind globally.

Please reflect on the enormity of it all – close to half a million (0.45M, now) images need to be processed - of restrooms, washbasins, urinals. In addition, it has to consider nuances in mobile image capture - size, color, shadow, lighting condition clarity, etc. Inasmuch, the design and the shape of objects including the surroundings aren’t of the same kind either. The complexity is compounded due to huge volumes, image diversity, the need for annotations/labeling to mark the boundary of the region of interest (commode area, floor area, etc.), and the subsequent review by business users.

Despite the numerous challenges arising due to inconsistency and large volume, the solution has an 85 – 90% accuracy (through a combination of deterministic & probabilistic approaches) with fraud detection capabilities and the governing principles of Ethical AI are adhered to.

The Solution Overview        

It has a mobile and desktop interface. The dashboards are intuitive and user-friendly for monitoring and tracking the cleanliness of the toilets across the education ecosystem, and can be used at the village level (by the Sarpanch), the taluka (Tehsildar), the district (The Collector), and at the State level as well (by the Education Commissioner, for instance).

An image clicked can be loaded with a unique code (school) and descriptors such as sub-category, student type (boys/girls), toilet kind, and positional value (latitude/longitude). After capture, it gives the user the option to “submit”, and once done, the rating is given – of the commode, floor, and overall. If the image is unclear or unsuitable, the ML engine nudges the user for better capture. The user can tag the images to the right school/ block and geo coordinates.

The Process follows a three-step procedure:

• Image Validation – Ascertains the given image - toilet seat/washroom/washbasin, or else it’s an invalid image.
• Object Detection – Whether the given image is of an Indian/European commode, or a washbasin, or floor area, or both.
• Sanitation Classification – Whether the identified object falls in the “Clean” or “Unclean” category.

The Solution Architecture

The AI Models are predominantly TCS’ IP Assets based on Open Source libraries of OpenCV, Python and Deep Learning, and Machine Learning models. It has advanced intelligent AI-based processing capabilities such as Image Pre-processing, Processing, and Post Processing components. Both On-premise & hybrid cloud versions are available.

The Hybrid cloud version runs on AWS, and can enable the following functionalities:

• Amazon EC2 - provides a wide selection of instance types optimized to fit different use cases. Instance types comprise varying combinations of CPU, memory, storage, and networking capacity that provide flexibility to choose the appropriate mix of resources for applications.
• Green Unicorn - "Gunicorn" - is a Web Server Gateway Interface (WSGI) that is commonly used to run Python web applications.
• CloudWatch provides data and actionable insights to monitor applications, respond to system-wide performance changes, optimize resource utilization, and provide a unified view of operational health. 
•  AWS Key Management Systems (KMS) - creates and manages cryptographic keys and controls their use across AWS services. This takes care of the audit trail, compliance, and fights fraudulent capture, etc.
•  GPU – Graphics Processing Unit on Cloud.

Responsible AI

A sensitive topic as toilet image capture has ethical considerations that need to be addressed mandatorily. Here are ways in which the standards are met:

• Explainable AI: Through audit trail of the entire decision-making process and inclusive of the decisions linked to the grading of the toilet’s cleanliness.
• Ethical AI: There’s fairness in processing the toilet images (as these are anonymous images).
• Performance of AI: The focus is on the highest level of reliability.
• Privacy of Data: Anonymous data – There is NO visibility as to who used the toilet and when, or whether the toilet was left uncleaned after use.
• Human-Centered AI enables higher efficiency in the cleaner’s job as the AI solution highlights what needs immediate attention.

Harnessing Intelligence@Scale

• SOA – Service Oriented Architecture lends portability.
• Dashboards designs are intuitive.
• AI@Scale due to non-linear Design Thinking

Simple touchpoints.

• Scalable – 45K schools & 0.5 million images can be processed daily.
• Sustainable through Digital Governance.
• Speed – portable & cloud-ready.
• Transparent – audit trail.
• Market Expansion possibility – 29 States & 7 UTs in India; 47 Least Developed Countries.

The Larger Picture – In Summary

The first of its kind globally, it also aligns with several UN Sustainable Developmental Goals. Among these are, Good health & well-being (Goal #3); Clean water & Sanitation (#6); Quality Education (Goal #4), and Reduced inequalities (#10). A nation can grow and transform if the talent pipeline is strong. If 1 in 4 girls are dropping out because of poor sanitation then this massive challenge needs to be solved as a priority. Moreover, it’s a global problem and such solutions can be mirrored in many other parts of the world. 

Image : TCS Clean India,Clean Schools