Chengbin Shou

Veniceland | Venice, Italy | Unit 21 | 2020

Click to see 4th Year work

Awarded Distinction for Thesis

Venezianella Castle, Veniceland 2020

A series of problems, such as natural disasters, over-tourism and self-contradictions, have posed massive threats on Venice. Countless politicians, scholars, and civic organisations have tried to save this historical city, but the future looks bleak. After rejecting the conventional solution, I assume that Venice will be owned by Disney in 2020 and transformed into Veniceland to save it.

Every Disneyland has a castle on its central axis. Therefore, this building project is the Disneyfied Castle of Veniceland. Inspired by Marinetti’s novel Venezianella e Studentaccio, I conceived the castle by using the facade form of the famous Venetian buildings and used CycleGAN (Generative Adversarial Network) algorithm to train the tool to transfer the style of the building facade and Disney cartoon characters. This unique architectural form merges the images of Venice and Disney. The huge castle named Venezianella will serve as the entrance to Veniceland, located at Lido inlet of the Venetian lagoon. In addition, the castle would become the monument of Veniceland and create the start of Veniceland’s narrative.

1. Venice or Veniceland?

Venice is a dying city not only because of floods, but also the over-tourism problems. In the ten years from 2007 to 2016, the number of tourists arriving in Venice increased by 38.8%. Compared with the top eight most visited cities in Europe, Venice has a smaller city-scale. The Airbnb phenomenon has become a challenging problem for Venice, and more and more primary homes are becoming second homes. To make matters worse, the Italian government has also been selling the federal property to the public. Venice is becoming a theme park for global tourism.

Out of concern for Venice, countless people sought solutions for these problems, which, in turn, has created two opposing camps. The first camp regards Venice’s Disneyfication as being representative of its tourism and one of the culprits destroying Venice; the other camp feels that Venice will have a smoother development path if it is actually owned by Disney. British economist John Kay is one of Veniceland’s strongest supporters. He argues: ‘If Venice were owned by the Disney Corporation, Venice would not be in peril.’ This view became the original concept of this project. The historical connection between Venice and Disneyland make this point more convincing.

2. Venezianella Castle

The themed castle is a key symbol of a Disneyland. Therefore, I set the castle in Veniceland on the artificial island of MOSE Project. MOSE (MOdulo Sperimentale Elettromeccanico, Experimental Electromechanical Module) is a project intended to protect the city of Venice, Italy, and the Venetian Lagoon from flooding. And the castle serves as the entrance of the Venetian Lagoon and the waterbus station, which prohibits giant cruise ships and manage the lagoon’s environment strictly. In addition, the castle would become the monument of Veniceland and create the start of Disney’s narrative.

Through the analysis of sight, it can be concluded that in order to retain the original landscape of the historical centre of Venice, the height of the castle needs to be limited to 315m or less. In order for other islands in the Venice Lagoon to see the location of the castle from time to time, the height of the castle needs to be above 135m.

Based on Filippo Tommaso Marinetti’s novel Venezianella e Studentaccio, the castle building a New Venice with the construction of an enormous vertical sculpture, using the architectural elements of famous Venetian buildings, which is similar to the design methodology of the Disney Castles. According to the building massing decided before, a collaged castle could be generated.

Click to zoom ->

3. CycleGAN

In order to retain both the features of Venetian architecture and the anthropomorphic style according to Marinetti’s novel, this project uses a machine-learning algorithm called CycleGAN.

CycleGAN is an approach to training image-to-image translation models using the generative adversarial network, or GAN, model architecture. This method is very powerful in style transfer and object transfiguration. It involves the automatic training of image-to-image translation models with two generators, two discriminators and two unpaired image datasets. With two unpaired image datasets, CycleGAN can train itself to decrease the feature loss of images step by step. In the end, the generated images could remain the structure and get the style transferred.

The training here uses the image collection of facades of famous buildings in Venice as dataset A, and the image collection of Disney cartoon characters as dataset B. After 200 training sessions, the resulting parameters can be used to transform Venetian building features and Disney character styles in both directions.

Click to open booklet ->

4. Detailed Design

By subdividing the facade of St. Mark’s Basilica to different degrees and then transferring the style, CycleGAN can generate disneyfied facade on different levels. Through a series of tests, the collage castle in this project can transfer its style with the help of CycleGAN training results.