Digital Camouflage Simon Weckert AI Evading Clothing
Simon Weckert made Digital Camouflage in Latvia using 65% recycled polyester, featuring seamless AdvTexture to disrupt AI object recognition.
Stock photo for illustration only, not from the actual event
- Digital Camouflage is a conceptual garment collection by Simon Weckert
- Textile patterns are designed to disrupt AI object-recognition algorithms
- Addresses fabric folding issues with a continuous, seamless AdvTexture
- Manufactured in Latvia using 65% recycled polyester and 35% polyester
Digital Camouflage is a conceptual garment collection by Simon Weckert that examines visibility, anonymity, and digital surveillance in public spaces. The garments are designed to interfere with artificial intelligence systems that use computer vision to detect people. At first glance, the clothing appears to feature an abstract graphic pattern, but the textile design is specifically developed to disrupt object-recognition algorithms used in surveillance, security, and autonomous systems.
The pattern is based on the concept of an adversarial attack, where visual data is manipulated so that a machine misinterprets what it sees. When worn, these garments make it difficult for AI-based person detectors to recognize the wearer, disrupting the system's ability to identify the human figure.
In computer vision, adversarial attacks often involve subtle visual noise that looks like normal patterns to humans but causes massive confusion for neural network models, leading AI to completely misclassify objects.
Previous attempts to evade AI person-detection systems relied on fixed printed patches, which became ineffective when fabric folded or camera angles changed. This vulnerability is known as the segment-missing problem. Simon Weckert's Digital Camouflage addresses this issue through a seamless Adversarial Texture (AdvTexture) covering the entire garment.
The continuous pattern is generated using a specialized generative AI method known as TC-EGA, which optimizes a tileable textile design. Applied across the full surface, the texture produces high-frequency visual noise and false visual features from different viewing angles, disrupting the consistency required for AI object recognition.
Stock photo for illustration only, not from the actual event
The collection is manufactured in Latvia using a durable blend of 65% recycled polyester and 35% polyester. The adversarial texture is applied through digital textile printing, combining computational pattern generation with textile production to challenge machine-based visual recognition systems.
Source: designboom
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment