CV based cyclist recognition
AI based bicycle and helmet recognition, the project from A to Z
Abto Software has entered into cooperation by an innovation provider catering businesses and governments. Their objective is handling regional challenges by delivering infrastructure services, IoT products, Smart City, and custom business solutions.
Our team has joined the partnership with the local provider to design a product for smart cyclist recognition. The project has resulted in the successful delivery of a demo-version solution for intelligent cyclist recognition, which can monitor bikers and non-bikers, exact speed, and helmets.
CV-based, smart cyclist recognition: Main goals
The client’s original mission was popularizing bicycle culture and construct an efficient cycling infrastructure. The project’s main objective was implementing an algorithm for accurate object recognition and monitoring, enabled through artificial intelligence, in particular computer vision.
Our team was involved to deliver a system that could:
- Track biker and non-bikers, exact speed, and helmets
- Ensure unprecedented recognition and calculation accuracy
CV-enabled, intelligent cyclist recognition: How the solution works
By leveraging domain-specific expertise, we designed a product helping collect and analyze traffic information. This includes detailed information about cyclists and non-cyclists that enter and exit the track, as well as some other details, for example if cyclists are wearing head protection.
Beside enabling object recognition, we ensured the solution can be also utilized to receive instant notifications. This way, the authorities can monitor those bikers not wearing head protection and take corresponding actions to encourage better adherence to acknowledged public guidelines.
How does this work?
A camera is placed right above the track to monitor both bikers and non-bikers that enter and exit that track. The video is streamed to an equipped workstation with a CPU/GPU hardware of sufficient processing power. That data is analyzed by the processing unit, which runs the system empowered through artificial intelligence. The system, previously trained and adjusted, quickly recognizes traffic participants to differentiate the cyclists from non-cyclists, and captures additional information, including velocity.
Our contribution
During prototyping, we covered:
a. Dataset preparation
At the first stage, to gather traffic information, a camera was placed directly above the multi-lane bicycle track. The camera was capturing both bikers and no-bikers, as well as identifying whether they’re wearing helmets and at which speed they’re moving.
b. Algorithm training in three successive stages:
- Biker and non-biker detection
- Helmet and haircut differentiation
- Speed recognition
The challenges
Accessing necessary data sets
Due to some delays on the client’s side, we had considerable trouble obtaining required data sets.
Setting equipment
As the company’s employees were facing significant issues with adjusting the angle and achieving high quality, we provided detailed guidance on how to install the equipment.
Choosing the most suitable AI model
Some models we considered are better for accurate speed recognition and others for precise helmet detection. To obtain desired results, our engineers gradually adapted and trained the selected AI model to meet the set business objectives.
Recognition nuances
The model sometimes confused voluminous hairstyles with helmets, accordingly disfiguring overall precision. To obtain required accuracy, our team took time for some additional work on training the adopted AI model.
Summing up
Abto Software has proven the feasibility of further software development to discover additional opportunities. Our client can provide the solution as a SaaS product to leaders, in particular to businesses and governments, that specialize in improving public surveillance and safety.
The project has succeeded in providing:
- Competitive advantage through providing a unique, innovative service
- Public surveillance and safety through monitoring cyclist behavior
- More interest in the cycling culture
- Enhanced reputation and image of the client company
