Global 
United States 
It is a well-known fact that the global population is on the rise alongside urbanization. This is in turn leading to the generation of huge heaps of waste matters of all forms, which are not being disposed sustainably through traditional methods. To tackle this problem, various innovative solutions have been considered such as AI-based recycling robots, Solar-powered trash compactors, AI recycling apps, e-waste kiosks, Pneumatic waste pipes, Waste level and garbage vehicle weight sensors, Plasma gasification, IoT-based streamlined waste collection and many more. But the most important checkpoint of waste management starts from the disposers, who can aid in waste segregation and recycling through automatic waste sorting.
AI assisted smart trash cans revolutionize waste segregation and disposal techniques, by automating the initial waste sorting process, necessary for innovative recycling solutions, thereby helping both the Government and the communal residents. Installation of these smart trash cans is being widely accepted, where this segment’s current global market value is USD 278.8 million, which is expected to increase to USD 5.42 billion by 2025, surging at a CAGR of 64.1%. [Frost & Sullivan]
The above graph shows increase in smart trash can market growth from 2015-2025 in North America
Unveiling the Potential of Smart Trash Bins
There are numerous challenges associated with waste disposal and management. For starters, the waste generated per day is a humongous amount of 637.99 billion Kg (807.58 billion people times 0.79 Kg per person per day). Then there are other issues such as health concerns, accumulation of pathogenic vectors, chemicals leaching into underground water table causing harm to aquatic life and environmental imbalance, animal loss due to consumption of plastic and other harmful material etc.
Traditional methods of waste disposal do reduce the overall amount of waste, but on the other hand, raise great concerns. Burning of non-biodegradable waste causes greenhouse effect, increasing global warming, melting of glaciers and water levels, thereby causing floods. Some other traditional methods include burial of solid waste increases traces of leachate in groundwater, dumping into water bodies (Sea Barging) causes aquatic life loss and rise of plastivores [Notable Instances: Great Pacific Garbage Patch, Lakeview Gusher], open landfills and waste incineration leads to greenhouse effect and the aforementioned issues, hog feeding and waste ploughing causes animal loss etc. Bio-degradable waste can be easily dealt with through bio-methanation and composting, but many a times, disposers fail to segregate such wastes due to improper labelling of bins or guidance.
Waste stream contamination is one major area of concern for the recycling industry. It basically refers to contamination in the batch of recyclable waste that was destined to be converted into new material. For example, solid material waste stream including paper waste gets mixed with glass, electronic waste or discarded food, thus, making it difficult to be disintegrated separately and directly converted into pulp for producing recycled paper, increasing processing time and labor costs. A small initiative by the user to properly discard waste in its designated place in separated trash cans can avoid such issues.
This is where AI assisted smart trash cans come into the picture. These smart bins utilize AI and Computer Vision models to detect and classify the waste item located in front of them and guide the user through an animated user interface or rotating specific bin lid or illuminating a specific bin or all of them to dispose waste in the right can, that is either red (unrecyclable household waste), green (food scraps and garden organics), yellow (plastic, metal containers, paper, cardboard) or blue (glass containers), thus, aiding in cleaning up the urban landscape through proper waste management.
As these cans automate the complete sorting process, long hours of manual sorting would not be needed. They eventually increase the overall recycling rate, reduce waste stream contamination by separating recyclable material at the beginning of waste management process itself, therefore improving recycled items quality and encouraging participation in such sustainable initiatives. Ultimately, with the help of automation, substantial municipal resources can avoid health- related issues caused during recyclable processing and can be allocated to stay focused on other important waste management aspects, thus, saving authorities spend.
How Do These Bins Work?
AI assisted smart trash cans are an amalgam of vision-based sensors, communication modules and AI processors that work with real-time data for providing efficient waste segregation support. Let us take a deeper look at the working of these cans
1. Human detection and pose estimation
Smart AI enabled cameras present in the AI enabled trash cans, detect the presence of humans in the scene, compute their depth/distance from the cans and also track them using onboard AI models. As the application tracks multiple humans in the scene, it also keeps analyzing their body pose, to detect any indication of pose or action meant for disposing the trash into the cans. When the required pose is detected, the person’s image and the trash will be processed by Image recognition AI models for trash detection and classification.
2. Image Recognition & DL Models
Cameras and vision sensors placed in these cans can detect, locate and recognize the waste item placed in front by using Image Recognition AI models that categorizes the trash based on visual characteristics like patterns, shape, color, size etc. Some separated trash cans also detect Radio Frequency Identification (RFID) and Near Field Communication (NFC) tags that define its overall material composition. These cans may also be equipped with volume estimation and load cell algorithms that help them sort trash items based on weight.
Altogether these cans categorize the waste material as recyclable (paper, glass, metal), non-recyclable (plastic, adhesives) and organic waste using Deep Learning models that differentiate sensor data with reference database containing information on various types of waste material and run decision making algorithms for segregation.
3. Actuators
Once the waste item is categorized, the IoT-powered sorting mechanism involving the bin lids open at certain angles automatically, indicating disposal in that specific bin. Some other types of sorting machines also work in a similar manner where conveyor belts or mechanical arms get activated and the belts carry waste items to separated trash cans for different types of waste material based on their composition.
4. User Feedback
Alongside triggering of actuators, these automatic recycling sorters provide users with instructions for disposal into specific compartments through visual text messages, animations and audio cues. In some cases, simple LED lights change colors to indicate the type of waste and its respective disposal compartment.
5. Active Learning
Smart notifications and diagnostics of each bin are communicated in real-time to the central management system authorities and service providers, when errors related to filled cans malfunctioning, maintenance or wrong sorting are encountered. Data collection, logging and analysis of performance statistics helps in active learning that includes saving new types of waste material which may not be present in database or are modifications or amalgams of base material and are of different quantities than reference database.
It also helps in improving user interactions through feedback loop algorithms as well as automatic waste sorting accuracy in real-world situations. These cans may be integrated with cloud–based platforms via Wi-Fi which aid in monitoring current recycling rates, trends as well as in remotely controlling configuration settings and sorting behavior.
The above image represents sample architecture of AI enabled smart bins
Join Hands with KritiKal
KritiKal Solutions carries technical prowess of over two decades in providing computer vision solutions and related software development. Its efficient team of experts developed an enterprise solution of AI assisted trash cans to classify the trash and guide the user to the correct bin. The vision pipeline consisted of accurate Deep Learning modules for detecting and classifying trash and even specific brands. It was also composed of the interactor module that conveyed information to and fro between backend and IoT sensors and actuators that lit up LEDs around the bin, displayed trash collection statistics, ads and stored trash count. It utilized Azure DevOps for remote deployment, monitoring, updation, and pushing diagnostics to cloud.
We have previously worked with some major clients in USA for developing customized self-sorting waste canswith intuitive user interfaces. You can turn your innovative and sustainable ideas into reality with us too. Please call us or mail us at sales@kritikalsolutions.com to avail our development services.
The Way Forward
Self-sorting trash cans present a promising sustainable future for human kind through competent and reliable waste management. These advanced technologies have been transforming our approach towards recycling for maintaining a harmonious environmental balance. Certain challenges and considerations such as high initial investment on technology and infrastructure, doubtful public adoption, technical limitations around sorting accuracy as well as necessity of regular maintenance and upgrades, may pose potential threats to its wider implementation under retail technology solutions, in malls, schools, colleges, parks, theatres etc.
With effective educational campaigns, efforts from municipal authorities for engaging community in sustainable practises, weighing of costs of installation and longevity against benefits and savings from these smart trash cans etc, imperative reduction in waste generation and methodical recycling can take place, therefore, supporting the overall implementation of a better waste management infrastructure and a cleaner future.
Yadhunandan heads the Vision Systems Design and Development function at KritiKal Solutions. He has an extensive experience of 19+ years in leading teams for research, development, product development, deployment and delivery of solutions using Deep Learning, Computer Vision and Machine Vision, for clients across different industries. Alongside being an IITian, he has successfully achieved 9+ granted patents and 4+ publications. He has dedicatedly helped KritiKal in various project deliveries to some major clients.