Jiangyue He

Jiangyue He is a womenswear designer from China. She aims to arouse the audience's attention and reflection on some social phenomena in reality by combining physical fashion and digital fashion. Jiangyue focuses on the problems faced by women in different work conditions. She hopes to challenge the audience's traditional perception of garments within existing markets.

In the digital age, artificial intelligence is widely used in the review of network information. Through the process of machine learning images, Jiangyue wants to let the audience see: where is female nudity most severely scrutinized in the eyes of AI? Are these places different from what we thought they would be?

As human beings, we can reflect on whether it is reasonable and necessary for us to censor women's bodies in movies, the Internet and normal life. Can there be a new definition of female’s nudity?

Jiangyue was very interested in artificial intelligence in the digital world. In the process of exploration, she mainly used the GAN language in computer code to express her theme. Before designing the final garment, she thought of herself more as a person who makes images. What she ultimately wanted to show was not only the garment, but also a lot of GIFs images.

During the final year in RCA, although she eventually did the work of fashion design, she positioned herself as an artist during the whole process. Seeing in her experimentation, garment and eventual repetition work, Jiangyue hoped her audience can not only have a rethinking of the new definition of female nudity, but also the absurdity of censorship.

Jiangyue's initial inspiration came from her experience of buying sanitary pads. The unnecessary thoughtfulness of the shop assistants made her feel ashamed of her female identity. This experience made her feel that being a woman in China is more closely scrutinized in society. Censorship is like an overly considerate shop assistant, constantly hiding what the audience already knows. As customers, we humans are perfectly free to laugh loudly at shop assistants. Or, we can change shop assistants and tell them it's not necessary.

Research about Internet censorship factory.

Experiment in C4D

Get myself censored in digital world.

scale=0 for image1

scale=0 for image2

In the process of AI training, my work content is mainly divided into two parts:

The first part is to train the AI with one identical image. During the training process, I adjusted the scale of AI generator. The smaller the scale, the less intelligent the AI, and it made more mistakes when producing the same image. Since the AI learns from the features of the image, the place with the largest range of change is the place that the AI thinks best represents the features of the image, and also the reason why the image is the most easily filtered.

— The bodies created by AI don't physically exist in the real world.

Female nudity created by AI.

In the process of AI training, my work content is mainly divided into two parts:

Unlike the previous part where the AI learned from a single identical photo, in this part I trained the AI with a large number of different photos and asked it to generate images of bodies that didn't physically exist, thus rendering the censorship meaningless.

— In the form of garment, the female body parts are extended to create visual association for the audience(Based on AI). I want to exaggerate some parts of the female body through fashion design, or make them irregular, so that while showing a natural beauty, the audience can reflect on the dispensability of social censorship of the female body.

Changes comparison between human model and AI model.

As I used a lot of printing of AI images on the garment, the AI nude models and human model existed at the same time in the whole shooting interface. Therefore, I put the photos back into the code in Stage2 to show how much the AI knew about the human model versus the AI model by how much the images changed.

In all GIF images, the AI model and the body parts of the human model wearing the pattern showed greater changes, indicating that the computer thought these areas best represented the image's features. These areas are not the faces and bodies of human model.

The comparison results are consistent with what I have encountered in reality. In communicating with my collaborators and photographers, the AI censorship system filtered out the AI images even though they were computer-generated. This makes it difficult for me and my collaborators to communicate with each other.

This result caused me to reflect on my whole subject. Because I use GAN to generate fake body images in the hope that the generated images will not be censored. In fact, in real life, eventually the AI censors the images that the AI itself produces, so my whole project becomes a circular, contradictory story. It makes no sense. It's ridiculous.

Jiangyue He

Social

About

Royal College of Art