The Bartlett
B-Pro Show 2021

Explore

About the show

0 other visitors

Mute Unmute Exit

unit-code

Project sections

1 / 1

Landscape Intelligent Value Enhancement System

Project details

Programme

Research Cluster RC20

While land makes up only 30% of the world, only 15% of it is protected, leaving 75% open to human exploitation. Furthermore, existing landscape protection legislation is inherently bound to an anthropocentric reading of nature. The two most common forms of protection – aesthetic uniqueness and ecological value – are human-imposed criteria. How might such laws and human-centric views be used towards non-human ends? Landscape Intelligent Value Enhancement System (L.I.V.E.S.) proposes a system for enhancing land value such that it falls under one or both of the aforementioned criteria.

Machine learning is used to decode various distinct landscape features from around the world, while generative adversarial networks (GAN) are used to generate possible outcomes. The various possibilities are first tested virtually against different ecological and environmental simulations. They are also turned into virtual tourist destinations that can be digitally photographed in order to evaluate aesthetic appeal. Modifications are then recoded onto the existing Arctic landscape, where the latent qualities of the selected site are thus enhanced and its ‘value’ increase in an attempt to flip the status from unprotected to protected.

Students

Decoding Protected Landscapes

Categories of IUCN Protection

Seven different protected area categories defined by the International Union for the Conservation of Nature (IUCN).

Database of Protected Landscapes

The ‘Money Shot’

For protected landscapes, the ‘Money Shot’ is the most impressive or memorable picture or scene. It is an integral part of the aesthetic value of a protected area.

Site Selection for Landscape Enhancement

Object Detection of Landscape Features

Recoding the Virtual Landscape

The Evolving Process of Landscape Layout

The latent space generated by StyleGAN.

Landscape Generation Strategy

Landscape generation starts with a plan layout that incorporates the morphology of the protected area. The topography of some unique protected areas is then used as a filter to enhance certain features.

Machine Imagination

Various virtual outcomes from StyleGAN.

Virtual Landscape Conversion

Combining subconscious awareness with intuitive responses, the system acquires human preferences for landscapes and transforms 2D images into 3D virtual landscapes.

Decoding Inherent Rules of Colour Combination in Nature

Using Pix2Pix GAN algorithm to decode the content of the protected landscape and recode them in generated virtual landscapes.

Self-Negotiation of the System in Decision Making

Choice of Location for Stamp Applications

The system applies landscape tiles generated by the virtual landscape to the locations with the best aesthetic potential.

The Generation Process of the ‘Money Shot’

Landscape Agent Simulation

With the help of machine learning agents, landscape agents are trained to generate the strategy for construction.

Building on Site

The Initial Stage of the ‘Money Shot’

The Evolving Ecology in the Arctic Landscape

Ecological Simulation System

Simulation Result for Vegetation Distribution

By simulating seed-spreading behaviour, the system is able to strategize the distribution of plants within the ‘Money Shot’.

Exploring in the Landscape

People enter the virtual landscape through a game-like environment. The system records the coordinates where the player takes the photo. Data is stored and compiled to determine the value of the ‘Money Shot’.

The Evolved ‘Money Shot’

Post-Anthropocentric Arctic

A brief description of how L.I.V.E.S. enhances the value of the Arctic landscape.

Share on Facebook, LinkedIn or Twitter

Index of Works

The Bartlett
B-Pro Show 2021

30 October – November 13

Explore