Today, Numina works in dozens of cities on three continents, and the demand grows as communities take more aggressive approaches to fighting climate change, recovering from the pandemic, and increasing mobility access.
As we deploy in diverse new environments, and as the mobility landscape innovates and adapts at an increasingly fast pace, we are tasked with measuring more forms of mobility and serving exciting new use cases, for example:
- We could not accurately measure streets in Kuala Lumpur without detecting motorbikes, and developing that capability laid the groundwork for us to detect additional modes of transportation more quickly in the future.
- We are helping an ever-growing list of cities measure multi-use patterns in public gathering spaces such as parks, trail systems, curbs, and Slow Streets.
- We are now serving transit agencies, universities, and industrials to understand specific interactions with curbside infrastructure and facilities.
What is Calibration Mode, and how does it benefit customers?
Calibration Mode is a two-week period of increased data sampling and data retention, designed to improve sensor accuracy in new environments. As with all our data policies, this mode has been designed to collect the minimum viable amount of data to achieve long-term data accuracy and mobility goals, while preventing the collection of potential PII.
Calibration Mode allows Numina to fine-tune sensors to their specific deployment environment and build training datasets to improve, or develop new, detectors for different kinds of objects (e.g. scooters). Performing a calibration period also allows customers to collect accuracy benchmarks for their specific Numina deployment environment.
When should sensors be put into Calibration Mode?
Whenever a sensor is deployed at a new location, Numina will put sensors into Calibration Mode for two weeks to get a better understanding of sensor performance and accuracy benchmarks in this new setting. If desired and necessary for further performance improvements, Calibration Mode may be extended or re-engaged, as needed and only with explicit authorization from the customer.
Can customers opt out of Calibration Mode?
Yes, customers may opt out of Calibration Mode for new installations. We will not put already-deployed sensors into Calibration Mode without explicitly asking, so customers will need to request (or opt in to) Calibration on currently active sensors.
Is this a new practice for Numina?
Yes and no. We have requested authorization from customers for similar calibration periods in the past. We are now standardizing this offering for all new and existing customers to improve customer experience.
What should I expect after Calibration Mode is complete?
After your sensors complete Calibration Mode, Numina will provide you with the results of the calibration period, including:
- Recommended coverage area(s) within the sensor view where Numina is most accurate, so you can create accurate custom Behavior Zones for reporting purposes.
- Accuracy metrics by mode and by behavior zone for each deployment location.
- Common errors in the data (such as distance, glare, or occlusion) and customized recommendations to address or eliminate these errors.
As an example, Numina recently performed a two-week calibration period for a new sensor we installed in Long Beach, California, as part of the City of Long Beach Smart City Challenge. The sensor was installed at a busy downtown intersection to measure bicyclist traffic patterns and desire lines through the intersection. Based on the results of a two-week calibration period, we provided the City with initial accuracy benchmarks by mode and Behavior Zone at this sensor. We also identified sensor data errors that were specific to this location, such as two trash bins in the background that were sometimes misdetected as distant pedestrians, and a pattern of missed detections at night in a certain part of the sensor field of view due to lack of street lighting. Together, Numina and the City drew a coverage area Behavior Zone that excluded these error-prone areas but still captured the section of the intersection that the City desired for reporting purposes. This ensures that the city team will have cleaner, more accurate data for the duration of their pilot study.
Where can I find more information?