As artificial intelligence (AI) continues to grow in scale and sophistication, the use cases for AI are popping up in all sorts of unexpected places.
The newest team to join the AtomLeap High-Tech Accelerator, CleanAI, is combining computer vision and AI to address a universal problem: cleanliness.
We sat down with Alistair Cadman, co-founder and chief product officer, to get his thoughts on the tech and vision behind CleanAI.
In your own words, what does the company do?
CleanAI is the first platform for automated detection of issues in car interiors and other environments. We combine computer vision with other sensors and machine learning to identify dirt, rubbish, lost belongings, spills, and other hazards. This enables real-time alerts and targeted responses for cleaner and safer spaces.
What problem are you addressing and why is it important to address it?
The first application of CleanAI is in a solution for shared cars and corporate fleets. This addresses the issue of dirty and messy shared vehicles and provides operators with notifications after each ride. This reduces the need for manual checking of the vehicle interiors and enables cleaning when it’s required. This means that if a vehicle is not clean it can be temporarily put out of service and cleaners dispatched before the next person has a poor experience. And if drivers leave belongings behind, they can receive instant notifications to retrieve their valuables before they go missing.
Importantly in this solution for car sharing vehicle occupants are never filmed during their ride. Our sensors are only activated once the ride is finished and the driver has ended their trip using the car sharing app.
How did you come across this idea (here, also explain the technology behind it, please)?
We first thought of the idea as regular users of car sharing in Berlin and other cities. We noticed that sometimes the cars are really clean, and other times they are really dirty and messy. After looking into the reasons why, we realized that operators don’t have real-time information about interior issues in their fleets. And that checking the cleanliness of cars and processing lost belongings are manual, inefficient processes.
Because we have expertise in computer vision and machine learning, we decided to tackle this problem using our technical know-how. We began prototyping using hardware that we developed in-house as well as trying both open source and custom AI detection models, before settling on a combination that provides a high degree of detection accuracy in a broad range of lighting conditions.
From idea to action: what prompted you to found your own company?
We decided to found CleanAI after receiving positive feedback from car sharing operators and car makers in Germany. On top of validating the problem, providing live demonstrations of our prototyping to potential partners also gave us new ideas for different environments that we could apply our detection AI to.
Following this early start in car sharing, we are now in discussions to conduct Proof-of-Concept/Pilot projects with several companies for applications both in and beyond car interiors.
From action to market: any thoughts you may have on the market?
Specifically for our first application in shared cars, the two major obstacles have been hardware and perception-based.
Firstly, in terms of hardware and the timing of our product conception, an obstacle is that the current generation of shared cars and vehicles in corporate fleets don’t contain cameras directed at the interior of the car cabin. This meant that in order to prototype our detection AI technology, we needed to design custom sensor hardware (including mounting) for retrofitting to the car interiors. This obstacle (and cost associated with it) is lessening, however, as several OEMs have recently announced that they will be including cameras in car cabins as standard. This opens up the opportunities for us to apply our detection AI using the imagine pipeline from these existing in-car sources.
Secondly, privacy is a major issue in Europe generally and Germany in particular. Anticipating that this would be the first question from potential partners, right from the outset we designed our first solution for car sharing with protecting customer privacy in mind. For this reason, our custom sensor includes a privacy cover that obscures the camera lens while the car trip is in progress. This lens cover then opens to enable an image of the interior to be taken once the driver has locked the car and finished the ride using the app. In the future when using OEM installed cameras without lens covers, as well as when applying our detection AI in different environments, we plan to use other privacy protection measures.
Who are the people behind CleanAI?
Philipp Nowakowski is the co-founder and CEO. He is a leader in fintech and taveltech, with a background in software development.
Alistair Cadman is the CPO with a background in design and innovation consulting in Australia and Germany.
Michael D. Kostrzewa is a software architect and engineer with expertise in computer graphics, video processing, and embedded development.
Artur Baćmaga is a software engineer with expertise in computer vision, NLP and machine learning.
What are your goals for the AtomLeap High-Tech Accelerator program?
As we expand our focus to different verticals for our technology, joining an accelerator was a natural next step. And with AtomLeap’s track record in supporting Berlin-based AI startups, it was a natural fit.
We’re looking forward to applying the resources afforded by the program to invest more time in our startup and to move beyond prototyping and demonstrations to conducting POCs and pilots with partners in Q3 and Q4 of 2019.
We’re also looking forward to the mentoring opportunities offered by AtomLeap, particularly around business plan development and finding market fit.