Image Deep

Role: Lead Computational Designer

Using AI and ML to reinvent buildings

Challenge

The projects aim is to reimagine a historic building precedent in a contemporary context. Reinforcing design concepts, tectonic, materiality and ecology to create a temporal evolving framework.

The precedent that i wanted to reimagine was Nemasus present in nimes. The challenges for this building was to meet the needs of a society for constant transformation and build houses at low cost. In doing so Jean Nouvel described a good apartment to be as large as possible, which is able to convert.A good cheap apartment in a democratic sense. The surrounding townscape is a mixture of low-level housing and light thus it is a mid rise building.  There is also a tradition of living outdoors in Nimes. 

Process

1. Precedent under study: To be changed

The challenges for this building was to meet the needs of a society for constant transformation and build houses at low cost. In doing so Jean nouvel described a good apartment to be as large as possible, which is able to convert. A good apartment is flexible, able to convert. A good cheap apartment in a democratic sense

1. Precedent 2 selected to incorporate certain qualities.

LJ wu design project beyond the shell was a stab at solving social housing problems in London. She created a grid frame structure with unfinished architecture paths. This creates a free form space which is flexible and enables the user to participate in enhancing the design. The fabrication was a super quick and was 3d fabricated

Analysis of precedent 1. Highlighting positive and negative elements

Key take aways and changes i want to make.

  • external terraces should be privately owned, and activity based rather than used for circulation.
  • These terraces should be activity zones. spilling into one another. – communal
  • Allowing for people to connect with one another.
  • Temporal growth and development. – through modular constructivism
  • should celebrate ecology and be used for growing.
  • Low-income housing community, provide for themselves? – Growing fruits and vegetables

Precedent 2 analysis and break down. Selecting keywords and elements in terms of prompts.

Generating AI images: Merging of Plans.

Precedent 1 + precedent 2 = resulting changed plans with integrated qualities

Generating inspiration images using keywords extracted from precedent 2.

Precedent 2 keywords + own keywords = Inspiration images (Future temporal setting)

Designing conceptual module simultaneously.

  • Preserving elements of precedent 1 and incorporating elements of precedent 2.
  • Strategies inspired by AI generated images.
  • Building raised to incorporate easy access, mass is broken into smaller zones to create opportunities for terraces. Connection to the exterior and a staggered affect

Generating AI images:

  • Merging of Sections.
  • Precedent 1 + precedent 2 = resulting changed sections with integrated qualities

Generating AI images:

  • Iterating generated sectional Images with prompts.
  • (Modular constructivism, flexible, unfinished)

Generating AI images:

  • (Eco village, communal terraces)

Generating AI images:

  • Merging iterated sections and plans together.
  • Incorporating architectural tectonic qualities complementing the site.
  • So the best opportunity to create greater interaction is through topological variation. Terraced affect. This not only provides people with housing but also a connection to the outdoors, flexibility in circulation, visual connection

Image to model transition.

  • Using Displacement mapping to translate architectural form into Rhino model elements.

Generating AI images:

  • Further iteration of sectional image.
  • (Slanted terraces, Complexity, expansion, variation)

Mid year review:

  • Designing and bringing parts together.

Rendered systems

Simultaneous Material explorations

  • Designing and bringing parts together.
  • Generating Prompts from existing images

Generating AI images:

  • Defining the construction system qualities.
  • Mapping material system to sectional image

Development after Mid review:

  • Integrating material system into the building

Defining the construction system qualities. Mapping material system to sectional image

Solution

A reimagined building concept which is created, managed and iterated through AI keeping in mind building surrounding context, communal housing concepts and novel material studies.

Impact

  1. Users can train their ML models so it can understand real world concepts
  2. Helps user test out and explore Novel and new ideas by layering methods.

Key Tech Stack