NVIDIA has developed a mind-blowing image reconstruction tool that can intelligently fix photos that are damaged or have holes and missing pixels.
The official blog post released this week describes it as “image inpainting” process that can be used in photo editing software to fill up the missing portion of images with realistic computer-generated pixels.
It is based on a deep learning system that can even fill up blank or corrupted spaces. For example, if a portrait has a missing eye, the AI system knows how to insert one even if the complete eye area is obscured.
“Our model can robustly handle holes of any shape, size location, or distance from the image borders,” said the developers of the image-retouching tool.
“Previous deep learning approaches have focused on rectangular regions located around the center of the image, and often rely on expensive post-processing,” added the researchers. But this model can easily handle holes of increasing size.
NVIDIA trained the system by generating thousands of random streaks and holes of arbitrary shapes that help the neural network to learn how to reconstruct photos intelligently.
Even though the results might not always be perfect, especially if the hole in the image is so large that there isn’t sufficient information for the system to reconstruct the pixels accurately.
But the system can prove to be incredibly useful in saving hours of labor, especially where the images are corrupted to the point where you need to generate whole segments right from the scratch.