SERVICE SHOWCASE
Imagetwin Launches AI-Generated Image Detection Capability, Strengthening our Joint Commitment to Image Integrity
Swapnil Midha
Head of Communications
Over the past year, TNQTech and Imagetwin have partnered to provide a comprehensive image screening service that detects duplicated or manipulated scientific images using a hybrid of technology and expert analysis. In a major step forward, Imagetwin has launched a new feature that detects AI-generated or AI-altered images, an increasingly urgent challenge in scholarly publishing.
How it Works
“AI-generated images leave behind invisible traces in the frequency domain, undetectable by the human eye. We have figured out a way to use specialised neural networks trained on tens of thousands of real and fake scientific images to detect this,” says Patrick Starke, CEO at Imagetwin.
“We generated thousands of AI images using image-to-image, text-to-image, and inpainting (a technique to modify specific parts of an image). We applied data augmentation like cropping, rotation, and scaling to these images during training to ensure robustness.”
The detection system is currently able to scan certain image types such as microscopy images, western blots, histology/pathology slides, cell cultures, and spot images. Future versions will expand to include plots, graphs, and other scientific visuals.
The new tool also supports bulk scanning, enabling publishers to check entire issues at once. Each flagged image comes with a confidence score. Images with high confidence scores can then be manually checked by publishers for further action.
Further reading
- Blog post detailing Imagetwin’s new feature
- TNQTech and Imagetwin’s service for the American Society for Microbiology
