Urban vibes

“Urban vibes” project is part of a larger project related to the study of the urban morphology in one of Barcelona’s neighbourhoods, Poblenou. With the aim of interpreting some of its key factors, it was interesting to use computer vision and machine learning algorithms to try to build a new dataset capable of suggesting information regarding the relationship between people and the urban environment.

</ p>
YOLOV5 QUALIFICATION

Object detection was used to observe and measure vibration at ground level across 4 categories:

– For the mixture of uses, bike lanes, occupation of cars, pedestrian pathways and other occupants of the street section, such as restaurant tables/chairs, were considered.

– For the existence of items; benches, trees, playgrounds and other specialized items.

– For the frontage activeness, observations were made on two sides of the street section,
separately and they are graded from highest (active in-between) to lowest (impermeable) activeness levels.

– For the mixture of activities; people’s behaviors have been observed mainly for walking, doing sports, standing and eating/drinking on the street.

</ p>
By using Yolov5, we were able to detect some of these parameters (people, bikes, cars, motorcycles, traffic lights/signs, umbrellas, chairs, benches, etc.) classifying them under three main categories: people, mobility and urban equipment.

</ p>
With the aim of generating a catalogue of comparable information able to relate the spatial characteristics of the road to its use, I used machine learning and image detection to extract this data from videos recorded on site.

</ p>
Using Yolov5 it was possible to count the number of objects present for each frame of the video.
In the second step to increase the detail of the recordable information I divided the video frame into 2 parts, left and right.

In this way, we were able to add a new level of analysis and understand the differences between the two sides of the street by relating them to the elements of the built environment, such as amenities, services, etc.

</ p>
MAPPING RESULTS

Once the video has been run through Python and Yolov 5.

We processed the information of the CSV files output from Python to Grasshopper, with the aim of mapping, visualizing and comparing the results with the morphological characteristics of Poblenou.