*In*Tent*City*Govhack 2019 Project “In Tent City”
We are a diverse team led by people experiencing homelessness. We’re using data to paint a portrait of homelessness in Victoria and to increase visibility on the services, where they are, what they are, the gaps in public data, and the surprising faces of homelessness.
Data on homelessness was drawn from AURIN, the Australian Bureau of Statistics’ 2016 Census, and the Australian Institute of Health and Welfare’s Specialist Homelessness Services data cubes and data tables on rough sleepers. Data on service demand vs. supply was drawn from Ask Izzy.
To estimate the difference between Google Images and reality, we downloaded a dataset of images from the search term “homeless melbourne”. We ran the images through a pretrained neural net to estimate age and gender. Then we compared the average age and gender of people experiencing homelessness in Google’s images to the numbers from the Australian Bureau of Statistics.
We used [https://github.com/hardikvasa/google-images-download] to scrape a dataset of the thousand top images for ‘homeless melbourne’. Then using [http://keras.io] and a pre-trained neural net [https://github.com/yu4u/age-gender-estimation] we estimated the gender and age of any faces in the images. You can check out the code at https://ethicalcode.github.io/intentcity/Estimate%20Age%20and%20Gender%20from%20Google%20Images.ipynb
Results: (graphs go here)