Jun 03, 2022
No image
Recognition algorithm for passenger count
Completed

Recognition algorithm for passenger count

$5,000+
2-3 months
Russia
2-5
Service categories
Service Lines
Artificial Intelligence
Software Development
Domain focus
Automotive
Consumer Products & Services
Other
Transportation & Logistics

Challenge

We trained algorithm to detect objects inside public transportation: adult passengers, children, fare inspector, personal belongings, etc. Space captured by the camera was divided into 3 separate zones(entrance, transit zone and exit) to determine passenger movement direction. Application allows to identify objects in transport at conditions (poor illumination, movement distortion) which are close to real environment.

Solution

Video is recorded by special tracking devices placed at the vehicle doorway. Video files can be transmitted online or at set period of time. We trained algorithm to detect objects inside public transportation: adult passengers, children, fare inspector, personal belongings, etc. Space captured by the camera was divided into 3 separate zones(entrance, transit zone and exit) to determine passenger movement direction. Passenger was considered to be boarding the bus if he initially appeared at entrance area and proceeded to the exit area. Consequently, passenger was considered to be exiting the bus if he moved from exit area to the entry area.

Results

Data labeling is taking significant chunk of the process which affects the result of neural network training. The more quality data there is, the more precise the recognition algorithm will turn out to be.
Data labeling is taking significant chunk of the process which affects the result of neural network training. The more quality data there is, the more precise the recognition algorithm will turn out to be.