Learn - How we are cleaning up our cities
Urban litter is a solvable problem. We are solving it with AI-powered vision and autonomous robots built for the real world.
The problem
Urban litter is one of the most visible and persistent environmental challenges facing our cities. Billions of pieces of rubbish end up on streets, parks, and waterways every year — creating public health risks, damaging ecosystems, and making communities less liveable.
Despite widespread awareness, the problem is growing. Current solutions rely almost entirely on manual labour — council workers and volunteers physically picking up litter piece by piece. It is slow, expensive, and unable to keep pace with the scale of the problem.
We believe the answer lies in automation. By combining advanced computer vision with autonomous robotics, we can tackle litter at a scale and speed that no human workforce can match.
Why it matters
- Public health impact
- Ocean & waterway pollution
- Wildlife harm
- Community wellbeing
- Tourism & city appeal
- Cost to taxpayers
Our vision model
At the core of our technology is a vision-language model (VLM) trained specifically to detect and classify litter in urban environments. We are building the most accurate model of its kind — capable of distinguishing litter from leaves, distinguishing a plastic bottle from a rock, even in poor lighting or cluttered scenes.
Training this model requires enormous quantities of carefully labelled real-world data. We collect footage from urban environments across Australia and beyond, and we work with volunteers to annotate every piece of rubbish frame by frame. The more data we gather, the smarter our model becomes.
This is an open research effort. We intend to publish our datasets and model weights so the broader community can build on our work and accelerate progress on this problem globally.
How you can help
- Label training data. Volunteer your time to annotate images and video clips through our training portal. Every label directly improves model accuracy.
- Share footage. Have a dashcam or phone footage of littered urban areas? We want it. Your contribution helps us capture real-world variety.
- Spread the word. The more people who know about our mission, the faster we grow our dataset and community.
The robots
Our robots are purpose-built for the urban environment. They need to navigate footpaths, avoid pedestrians, handle kerbs and uneven surfaces, and operate safely alongside the public — all while identifying and collecting litter with precision and reliability.
We are designing these robots from the ground up, informed directly by the capabilities of our vision model. The robot and the AI are co-developed: as the model gets better at seeing litter, the robot gets better at picking it up. This tight feedback loop is core to how we work.
Our first target environments are the streets and parks of Sydney, with plans to expand to other Australian cities and eventually international urban centres as our technology matures.
Design priorities
- Pedestrian safety
- All-weather operation
- Low noise
- Long battery life
- Compact form factor
- Remote monitoring
Our principles - How we think and build
We are an early-stage team tackling a hard, real-world problem. These are the principles that guide every decision we make.
- Accuracy first. A robot that picks up the wrong thing — or misses litter entirely — is worse than useless in the field. We obsess over model accuracy before anything else.
- Safety by design. Our robots will operate around real people in real places. Safety is never an afterthought — it is baked into every design decision from day one.
- Open by default. We believe this problem is too important to solve behind closed doors. We share our data, models, and findings openly wherever possible.
- Community-powered. Our vision model is only as good as the data behind it. We rely on our community of volunteers to help us build the richest, most diverse dataset possible.
- Built to scale. A single robot cleaning a single street is a proof of concept. We design our systems from the start to be deployable as a fleet, across a city, then across a country.
- Impact-driven. We measure our success in kerbs walked, pieces collected, and waterways protected — not vanity metrics. Every engineering decision is tied back to real-world environmental impact.