Improve recycling efficiency using
computer vision and machine learning

Concept Video

Recycling is an important component 1 in creating a sustainable and circular economy. Over the recent decades we as a global society have made some small steps in the effort to stop climate change, but the recent IPCC report points to the urgency of reducing emissions.

In addition to other efforts 2 to reduce emissions, it is now ever more important to recycle acurrately as cities are facing situations where much of the recyclables collected can end up in landfills since China now rejects plastics using a stringent threshold of contaminants . Articles such as 6 Things You’re Recycling Wrong can certainly help, but only reach a small percentage of the population 3. Having feedback at the point of disposal will have immediate impact on the accuracy and efficiency of our recycling infrastructure.

Prototype showing positive feedback when a person puts the correct item in a recycling bin


While people in many cultures over time learn and teach each other to recycle correctly, it is still an error-prone process. Many factors contribute to incorrect sorting of common products, for example:

  • The categories are often confusing, and organisations try to solve this with large complicated posters which not everyone reads
  • People can forget what they had previously learnt
  • People can be in a hurry or make absent-minded mistakes
  • People in some cultures are less diligent and disciplined than others
  • Visitors to foreign states or countries are unaware of local customs
Recycling containers by TriCiclos in Chile

Recycling containers by TriCiclos in Chile

Recycling mistakes by employees at a corporate HQ

Recycling mistakes by employees at a corporate HQ

City of Barcelona placing ads on public transport to educate people

City of Barcelona placing ads on public transport to educate people

Feedback is most effective at the time of disposal, at the site of recycling sorting and directly in response to the individual’s actions. Personally I have observed citizens hesitating while recycling and lazy tourists dumping rubbish in any receptacle. Whatever the reason such a system would have an effect on improving the accuracy of objects being placed in the correct containers, resulting in lower contamination. Even for people who don’t care, having immediate feedback could prevent mistakes and act as a teaching tool.

In some parts of the world creating sorting centres is not something that all countries can afford, as was the case for Chile. The efficiency of recycling in such cases are wholy dependent on how well the population recycle.

In addition to being a feedback device, the system can also serve as a useful tool in collecting data and statistics as it is an IoT sensing device.

  • Refinement of recognition system by error reporting and assessment of random samples
  • Statistics on how well residents or employees are recycling
  • Comparison of recycling rates of different neibourhoods of a city in order to target efforts more effectively
  • Statistics on what type of item are people most confused about
  • Estimation of how full a container is


The concept was conceived initially for city-level containers, but can work equally well for bins in malls, airports and offices or even smart homes. Devices can be adapted to target any site where sorting or separation of waste happens including both public bins and city-wide sorting facilities.

The prototype seen in the video was deployed onto a mobile phone, but devices suitable for mass deployment will likely be a small embedded device. City and region specific details will be loaded onto the device, such as locale-dependent recyclable categories, and software updates are pushed to it periodically. Visual feedback is customised to the particiular city which has a specific colour-coding scheme for different recyclables, and can be configured to display multiple languages to reach both locals and visitors alike.

Feedback mechanism can range from an LCD display, a projection to simple coloured LEDs. Audio cues can be added to reinforce the visual feedback but are unlikely to work well by themselves. A gamification angle, such as the award of points as positive reinforcement, can be explored to encourage effort and participation.

Prototype device reacting to various items

Prototype device reacting to various items

Example hardware set up using off-the-shelf components

Example hardware set up using off-the-shelf components

Next Steps

While the project is at a nascent phase there are no significant challenges that the current state of technology cannot overcome. Please support this project by reaching out to your local government or ecologically-minded organisation and telling them about us.

  1. The other big R’s such as reduce and reuse are obviously very important also. [return]
  2. For example, using renewable energy sources to power the electricity grid, in which both government and large corporations can do more. [return]
  3. People who click on and read an article on recycling are likely to be people who are already putting in an effort to recycle well [return]


Please email us if your public or private organisation would like to inquire about the project.

Contact Us

Follow us on Twitter

Become a friend on Facebook