AI Identification of Light Pollution Sources

Let's analyze light pollution using artificial intelligence (AI)!

Light trespass is particularly evident in densely populated cities like Hong Kong. Which light sources are most likely to cause light trespass? How are they distributed across different areas?

To tackle this problem, we have taken a large number of photos across different locations in Hong Kong. We developed artificial intelligence (AI) algorithms to detect, classify, and quantify light sources. With deep learning, these algorithms utilise zero-shot object detection and image understanding AI techniques to differentiate objects from the background. Based on shared features, they assign identified light sources to different categories, thereby enabling the recognition of the objects from which the light sources originate.

We are also developing a light pollution index which reflects the state of light pollution by measuring its direct impacts on residents and natural habitats. We aim to promote better lighting practices, enhance energy efficiency, and reduce light pollution.

As part of a citizen science project, we encourage the public to contribute images through this web app to expand the coverage of our study.