Machine vs Machine

Singapore blogger Daryl Aiden Yow created a storm when he manipulated stock images and posted them on his Instagram account. The 27-year-old has since apologised but his action has cast more doubt on the blogging community.

With a little knowledge of Adobe Photoshop, one can do wonders with an image, such as Daryl planting himself in exotic places.

However, artificial intelligence (AI)-enabled tools are taking image manipulation a step further.

For instance, NVIDIA researchers have come up with a state-of-the-art deep learning method that can edit images or reconstruct a corrupted image — even those with gaps or missing pixels.

Called “image inpainting”, the method could be implemented in photo editing software to remove unwanted content, while filling it with a realistic computer-generated alternative.

On the other side, Adobe is working on a counter-measure that also leverages the capabilities of machine learning to automatically detect edited pictures.

What used to take a forensic expert hours to do can be done in seconds, according to the Adobe blogpost.

The results of this project are that AI can successfully identify which images have been manipulated. AI can identify the type of manipulation used and highlight the specific area of the photograph that was altered.

This looks like another battle between machines — one that creates a fake image and another that detects the manipulated image.

Tagged with: