About the Algorithm
The Process
First, hundreds of pictures of liminal spaces were gathered off of different sources, from subreddits to fine art photography or 3D simulations. All these and counter-examples of non-liminal places were used to train the software.
The Tools
The algorithm uses a specific kind of machine learning (ML), more specifically a Bag of Features model to detect objects. It relies on the Vision API provided by Google. This AI analyses images and is able to detect elements (architecture, lighting, people, words, objects…) therein, which was the base for the liminality calculations.
The Calculations
For each element most frequently identified in images considered ‘liminal’ (by various online communities), a Coefficient of Liminality was calculated. It’s that data that allows the algorithm to gauge a Liminality Score to every given picture.
Images analysed by the algorithm:
2,866
And counting.
Submit a place
Send a picture of a location to evaluate its level of Liminality.
You can also add commentary with the image (i.e.: what kind of space it is, what context was this picture taken in, where is the picture from…)
Send an email at theliminalbot@gmail.com
© 2022 - vacuité | Powered by TM
