Excellent | SDG11 | SDG12 | SDG9 | Trash collection/ waste management | United States


Share this post

1. General


SDG 9: Industry, Innovation and Infrastructure

SDG 11: Sustainable Cities and Communities

SDG 12: Responsible Consumption and Production


Trash Collection/Waste Management

2. Project Details

Company or Institution




General description of the AI solution

Diwama is developing an AI-based image recognition software that automatically detects, classifies and tracks waste at multiple stages in the waste value chain. Waste is characterized based on type, weight, brand and quality to provide data and actionable insights for waste managers, organizations and governments thus helping them improve waste management operations and predict waste generation trends for better investments allocation.
For the project nominated, the AI algorithm is integrated on both the consumer and industrial levels. By integrating Diwama’s algorithm into a waste collector mobile application, we are helping people sort better at home by simply scanning the waste items and the AI will tell them if it is recyclable or not. This way we increase recycling awareness and sorting at source which in turn results in higher quality recyclables for the collector.
The collector then collects the sorted recyclables from the households and transports them to his sorting facility where Diwama’s algorithm is also integrated with a camera over conveyor belts. Waste is detected and tracked in real time based on what the camera sees. Data collected is then displayed on an online dashboard which helps the facility manager make data driven decisions to increase recovery and purity of recyclables. In addition, the already existing manual sorting is improved by putting a screen in front of workers with live detection of waste. The screen acts as a visual aid for the sorters where they can recognize the type of waste by a colored rectangle around the confusing items so they can sort them into the correct bin. Finally, through the brand detection feature of Diwama’s AI algorithm, waste is identified by brand and accordingly the brands are informed on how much of their products are getting recycled to help with their extended producer responsibility reporting.





3. Aspects

Excellence and Scientific Quality: Please detail the improvements made by the nominee or the nominees’ team or yourself if your applying for the award, and why they have been a success.

Diwama’s core technology is a deep learning algorithm trained to recognize billions of waste items in real time at 30 frames per second and accuracy above 90%.
Diwama’s product consists of 3 layers of deep learning algorithms:
1st layer: Optical character recognition that reads specific words and brand names on the waste items to make a preliminary classification of waste type. For example, Pringles is a non-recyclable tripak material (made of 3 different types of materials). Another example is reading the word sardine or tuna and directly associating the object with a metal can.
2nd layer: Yolo V4, an open-source algorithm, in combination with Onnx-Tensort RT to detect objects based on shape, size, colors, etc. Getting a similar result as the 1st layer of detection is confirmation of the object type. In case there is a difference in material detection, the image is taken into the 3rd layer of detection.
3rd layer: This layer is used as a final confirmation if the first 2 layers did not provide similar results and to act as an automatic training for the algorithm (unsupervised learning) to save on manual annotation and labeling of data. In this layer, the bounding box generated around the object detected through Yolo V4 is taken as a screenshot where the item is isolated from all the surroundings and an image classification algorithm is applied to classify the object. Once the object is classified, the final result is whatever it matched with, layer 1 or 2, and the layer that did not match is automatically relabeled and fixed to train it with time and increase its accuracy.
With the release of any new open source algorithm with higher performance, any of the algorithms can be easily replaced while maintaining the same solution architecture.

Scaling of impact to SDGs: Please detail how many citizens/communities and/or researchers/businesses this has had or can have a positive impact on, including particular groups where applicable and to what extent.

Through Diwama’s Vitron, waste sorting facilities are able to increase their recovery of recyclables by up to 15%; recycling facilities are able to increase their plastic pricing by up to 20%; and governments are able to measure waste generation trends and accordingly allocate investments where needed for adopting clean technologies.
Diwama’s impact on the short term is to divert additional 1.5% of waste from landfills in year 1, 2.5% in year 2 and 3.5% in year 3. This corresponds to an additional recovery of recyclables in facilities and thus an increase in recycling rates and profit margins. This impact reduces waste generation since more is getting recycled to address the SDG Goal 12 of Sustainable Consumption and Production, target 5, as well as the SDG Goal 11 of Cities, target 6, of reducing the per capita environmental impact.
Through Diwama’s solution, facilities will be able to increase their revenues by up to 13 USD per ton of waste in 5 years. As an example, for a city with half a million inhabitants, there would be a 500 tons per day waste sorting facility with a potential of 2.37 million USD revenue increase using Diwama’s software.
While this is not related to reducing leakage of plastics to the environment, an increased profit margin and throughput for waste management facilities will incentivize larger investments in the industry to ultimately reduce leakage into the environment and oceans. Thus addressing the SDG Goal 9, target 4, of upgrading infrastructure to make them sustainable with increased recourse-use efficiency.
Finally, Diwama is not aiming to get into robotic sorting as it replaces humans and reduces job opportunities. Diwama is focusing on improving the work conditions of existing manual sorters to increase their safety and productivity within facilities across the globe.

Scaling of AI solution: Please detail what proof of concept or implementations can you show now in terms of its efficacy and how the solution can be scaled to provide a global impact ad how realistic that scaling is.

Diwama is working with a small waste sorting facility in Lebanon with only 6 tons per day capacity. However, Diwama is currently planning to install in one of the largest facilities in Lebanon and MENA at a capacity of 1,200 tons per day. This is due to Diwama’s technology of processing 30 frames per second which makes it possible to detect waste at high speeds on conveyor belts.
Diwama is also piloting with Veolia Middle East in Abu Dhabi on a small sorting station as well as integrating the API of the AI algorithm into their mobile application to help citizens sort better at home by simply scanning the waste item through their mobile phone camera and the AI tells them if it is recyclable or not. Here the dataset of images used is completely different than the ones used for the industrial scale system.
Another early adopter is a company working on smart bins, where Diwama’s API will be connected to their cameras inside bins to detect the type and brand of waste thrown.
This diversity of data is allowing Diwama to build a robust, flexible and scalable algorithm that can be integrated almost in any application for waste detection. This connects all the players in the waste value chain; from waste generators to government officials and waste management companies, including sorting and recycling facilities, collection companies and others, to adopt this AI technology and improve their performance at all activities.
Currently, the product is capital intensive to collect and process data and it takes at least 2 months to install in a new application. However, the costs will reduce significantly at economies of scale and with more projects adopted, providing a profit margin of up to 50% for Diwama and only a 2-week installation period to any customer.

Ethical aspect: Please detail the way the solution addresses any of the main ethical aspects, including trustworthiness, bias, gender issues, etc.

Robotic sorting systems that sort waste based on AI have been in the market for a few years and are currently being scaled up. However, such technologies are very expensive and replace human sorters which is not an option in global south countries, especially that more than 2% of the population depends on the waste sector to make a living. In addition, global south countries face a huge waste management problem due to the gap in technology used in the developed countries, and through robotics this gap is increased. This is why Diwama is focusing on the software component, to bridge the gap between developing and developed countries, and thus help developing countries manage their waste in a more efficient and cost effective way without the requirement for large technology investments.
Diwama’s main value proposition is collecting waste data and using it for improving several activities across the waste value chain. All the collected data from the different installed systems are to be stored on the cloud in one place in a secure manner. No personal data will be collected but rather data on the macroscale of waste generation is collected. Finally, whatever data that should not be leaked or given to anyone is kept secure with no access to it other than the owner of the project.


International Research Centre in Artificial Intelligence
under the auspices of UNESCO (IRCAI)

Jožef Stefan Institute
Jamova cesta 39
SI-1000 Ljubljana




The designations employed and the presentation of material throughout this website do not imply the expression of any opinion whatsoever on the part of UNESCO concerning the legal status of any country, territory, city or area of its authorities, or concerning the delimitation of its frontiers or boundaries.

Design by Ana Fabjan