Public spaces in Canadian cities are often designed without adequate tools to measure and compare the risks of exclusion based on factors such as gender, age, and religiosity. As cities become increasingly diverse, marginalized communities—including elderly women, LGBTQ+ individuals, people with disabilities, and indigenous populations—experience differential access and utilization of these spaces. This lack of inclusivity can lead to socially fragmented urban environments and reduced social cohesion.
Semi-Structured Interviews:
Conducted with representatives of diverse groups in Montreal to gather insights into the utilization of street spaces by marginalized communities.
Focus Group Exercises:
Engaged 20 participants from varied backgrounds to evaluate aspects like safety and accessibility in urban spaces using curated Mapillary images.
Pairwise Image Comparison:
Pairwise comparisons and labeling of 15,000 Mapillary images, guided by criteria developed from interviews and focus groups.
Algorithm Training and Evaluation:
Fine-tuned a Multi-Layer Perceptron (MLP) pretrained on ImageNet to correlate image attributes with inclusivity scores, followed by workshop evaluations with participants.
The heatmap below illustrates the inclusivity of various public spaces in Montreal based on our AI model's analysis.
Explore the inclusivity heatmap here.
The heatmap generated from the AI model highlights areas in Montreal that are more or less inclusive. Key findings include:
-Platform Development: Create a user-friendly platform to facilitate public consultations, allowing citizens to visualize and contribute to the design of their neighborhoods.
-Data Expansion: Increase the dataset with more images to improve AI model accuracy and reduce biases.
-Partnerships and Funding: Seek collaborations and funding to pilot the platform and expand the project’s reach.