Image recognition, reconstruction – and how AI can go wrong

0
Showing 5 of 8

Privacy and pixelation

Source: Menon, Sachit, et al. "PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2020.

For many years, we have been protecting people’s privacy by pixellating their faces in images. However, GANs have blown that technique out of the water. They can reconstruct missing or blurred elements of images using a technique called inpainting. This includes being able to restore low-resolution images to a much higher resolution.

Share

About Author

Veronique Van Vlasselaer

Data science enthusiast with an infinite passion to operationalize business decisions supported by insights from data analysis, machine learning and AI; anti-fraud expert; team player who loves to work with inspiring people, main focus = having fun on the job

Leave A Reply

Back to Top