Seeing Cities Through a New Lens: How Tech is Measuring Urban Quality

Cities are more than just concrete and steel. They have a vibe, a feel, and a quality that can make or break how people live and behave. But how do you measure that?

The Study

A recent study used a mix of street-level photos and computer smarts to rate the quality of urban environments across the U.S. The focus was on five key aspects:

Beauty
Relaxation
Nature
Walkability
Safety from crime

Data Collection

Over 72,000 people ranked street-view images on these qualities.
These rankings were used to train deep learning models.
The models could predict the quality of streets at 120 million locations for the years 2008, 2012, 2016, and 2020.

Accuracy

The models weren't perfect, but they were better than random guesses, with accuracy ranging from 59% to 73%.

Challenges

Bias

One big challenge was bias. The study found that the models were less accurate for certain demographic groups, like:

Hispanic/Latino
Native Hawaiian or Pacific Islander communities

Even after adjusting for these biases, some gaps remained.

Seasonal Bias

The study also found that images taken in late spring and early summer scored higher in quality. They adjusted for seasonal biases too.

Results and Implications

The results give a nationwide snapshot of street-level quality. This info could be useful for:

Public health research
Urban planning
Policy decisions

However, the model for safety from crime didn't perform as well as the others, which could be a problem if this data is used for planning or research.

Conclusion

This study shows how tech can help us understand cities better. But it also highlights the challenges of bias and accuracy. As with any tool, it's important to use this data wisely and be aware of its limitations.