NVIDIA has introduced an amazing “state-of-the-art deep learning” method that can reconstruct an incomplete image. Last week the company released a video showcasing the examples of the new tool.
The main competitor for this new tool is Photoshop’s Content-Aware fill. It uses information from surrounding pixels to fill gaps in the image. However, this tool fails badly when it encounters a more complex image. NVIDIA took a different approach and put its new tool ahead of the competitors by employing AI technology that understands what the subject should look like. However, it remains unknown when this tool will be available for the consumers.
Even if this new method isn’t excellent at making esthetic reconstructions, it is definitely very good at recognizing what each part of the image represents. All the imperfections aside, we can’t wait to test it out.
Wes’ note: Consider how this might be used to fool us! Will we be able to depend on photos at all?
More info: news.developer.nvidia.com