Park Connector Network (PCN) is a system of greenways strategically planned to link parks and open spaces across the entire Singapore (Tan 2006). It functions both as nature corridors that effectively strengthen biodiversity and ecological resilience, and open spaces for people’s everyday life that help to enhance community resilience and social sustainability. However, literature on Park Connector Network largely centred on its ecological performance. Little is known about people’s daily use of the greenways, and how their activities are related to the physical environment.

This research aims to bridge this knowledge gap using cutting-edge deep learning technologies and big data analytics. First, a three-stage location-tagged video survey with GoPro Hero 5 and Canon 5D was conducted to capture people’ use of the PCN throughout the entire network. Their presence was plotted and geo-registered using object detection with a Mask R-CNN model (He et al. 2018). And the specific physical, social and recreational activities were identified and inferred using spatio-temporal action localization with models trained on AVA datasets (Gu et al 2018). Second, the physical environment of PCN was assessed at a fine-grained scale using semantic segmentation with a PSPNet model (Zhao et al 2017), which can detect and quantify up to 150 different objects such as sky, trees, buildings, chairs, lampposts, etc., based on a large number of panoramic images of the greenways captured with NCTech iSTAR Camera. A huge database was then constructed that enables in-depth examination of the correlations between environmental qualities and human activities, and identification of the most salient environment features on PCN usage.

These innovative methods for measuring, analysing and evaluating environment-behaviour relations potentially can help to inform decision making in the planning and design of future PCN and other green spaces in Singapore to further enhance its ecological and social resilience.