A brightly coloured Hawaiian-style shirt has been designed to confuse artificial intelligence (AI) surveillance systems, making it harder for certain cameras to recognise the person wearing it. Created by Berlin-based artist and technologist Simon Weckert, the garment is called Digital Camouflage.
Its unusual combination of colours, shapes and patterns is intended to disrupt the visual features that AI-powered cameras use to identify people. While the shirt may look like an unconventional fashion choice to human observers, its design has a specific technological purpose: to exploit weaknesses in computer vision systems used for automated surveillance.
How the Digital Camouflage Shirt Works
AI-powered surveillance cameras use computer vision technology to analyse footage and identify objects, including people. These systems are trained on large image collections to recognise visual patterns associated with human bodies, such as silhouettes, proportions, and the relationships between body parts.
Weckert's Digital Camouflage shirt is designed to interfere with this recognition process. Its busy pattern combines contrasting colours and overlapping shapes, making it harder for detection software to identify the wearer as human. Unlike humans, AI models rely on patterns learned from training images to classify what appears in front of a camera.
The shirt remains visible to the human eye but can cause certain AI detection systems to fail to register its wearer as a person. However, it does not make someone physically invisible or prevent cameras from recording them. Instead, the design targets the software analysing the footage, and its effectiveness depends on the detection system being used.
AI Was Used To Develop the Pattern
The shirt uses a technique known as an adversarial attack, which involves manipulating visual information to expose weaknesses in an artificial intelligence model. In this case, the pattern was developed to interfere with the features that object-detection software uses to recognise people.
According to the project's documentation, Simon Weckert developed and tested the pattern using YOLO-based object detection and an open-source system informed by research into adversarial textures. The approach aims to exploit how computer vision models interpret colours, shapes, and other visual features.
The shirt was tested against YOLO, an open-source object-detection model family. During demonstrations, the system failed to register the wearer as a person. However, the results do not establish that the garment can defeat other surveillance systems.
Designed in Response to Surveillance Plans in Berlin
Digital Camouflage was developed in response to plans for AI-powered video surveillance at Kottbusser Tor, a busy transport hub in Berlin's Kreuzberg district. The location has been a focus of discussions surrounding public surveillance and the use of technology to monitor activity in urban spaces. Weckert's project explores how clothing might challenge the automated systems used to analyse public areas.
Rather than relying on traditional camouflage designed to blend into a natural or urban environment, the shirt uses a pattern intended specifically to interfere with machine perception. The project raises questions about how AI surveillance systems interpret the world and whether deliberately altered visual information can affect their ability to recognise people.
The Shirt Has Limitations
Although Digital Camouflage can confuse certain object-detection systems, it does not guarantee protection against surveillance. Its effectiveness depends on the AI model being used, while factors such as camera angle, lighting, distance and movement may also affect how the software processes images.
The shirt is designed to interfere with object detection, which identifies people and objects in images, rather than facial recognition, which analyses facial features to identify or compare individuals. Disrupting one system does not necessarily affect the other.
Digital Camouflage does not prevent cameras from recording the wearer or guarantee that they cannot be tracked. Its demonstrated effects are limited to certain detection systems.
A Different Approach to Anti-Surveillance Clothing
Digital Camouflage is part of a wider range of design projects exploring how clothing can interfere with surveillance technology. Other approaches include garments fitted with infrared LEDs to affect certain camera systems and printed patterns intended to challenge facial recognition or object-detection algorithms. These designs explore the relationship between fashion, privacy and automated monitoring.
Weckert has previously examined weaknesses in digital systems through his artistic work. In 2020, he used 99 smartphones carried in a handcart through Berlin to make Google Maps display a traffic jam where no actual congestion existed. The experiment demonstrated how simulated digital activity could affect the platform's traffic information.
His Digital Camouflage project applies a similar approach to AI-powered surveillance, using clothing to demonstrate potential weaknesses in automated visual recognition. However, the Hawaiian-style shirt does not offer universal protection against surveillance, and its effectiveness depends on the technology analysing the images.
