Google’s AI-based upscaling technology can generate high-resolution images from poor quality images
- The two diffusion models can generate high fidelity images.
- These models have plenty of applications, including the improvement of medical imaging systems.
AdvertisementGoogle has released AI-based image upscaling technology that is said to enhance the quality of low-resolution images. In a post on Google’s AI blog, the researchers from Brain Team introduced two diffusion models to generate high fidelity images. The two models are image super-resolution (SR3) and cascaded diffusion models (CDM).
What is Image Super-Resolution?
First of this model is the image Super-Resolution via Repeated Refinement or SR3. This method is defined by the research team as “a super-resolution diffusion model that takes as input a low-resolution image, and builds a corresponding high-resolution image from pure noise.”
To implement it, the machine uses a process of image corruption where noise is consistently added to a high-resolution image until only pure noise remains. It then reverses the process that removes the noise and reaches a target distribution “through the guidance of the input low-resolution image.”
Some of the results shown by Google’s research team are impressive and showcases how this method can be used to effectively improve the image quality of low-resolution images. As per the post, the super-resolution can have multiple applications that include enhancing the existing medical imaging systems and restoring old family portraits.
Cascaded Diffusion Models (CDM)
Once the SR3 model had shown effectiveness, the Brain Team used the model for class-conditional image generation. CDM is explained by the researchers as “ a class-conditional diffusion model trained on ImageNet data to generate high-resolution natural images.”
As per the post, Google built CDM as “a cascade of multiple diffusion models” since ImageNet was a difficult, high-entropy dataset. The model is a combination of multiple diffusion models that can generate images of increasing resolution. It starts with a standard diffusion model at the lowest resolution and is followed by a sequence of super-resolution models that can successively upscale the image and add higher resolution details.
Today we present connected approaches that push the limits of high-fidelity image synthesis through use of a pipeli… https://t.co/tCkynbHusa— Google AI (@GoogleAI) 1626459611000
Along with SR3, Google also uses a new data augmentation technique, called “conditioning augmentation”, that is said to further improve the sample quality results of CDM.
Using the CDM method, a low-resolution image of 64x64 can be diffused to 264x264 resolution and then further to 1024x1024.
With the introduction of these models, Google is looking to improve the natural image synthesis that has wide-ranging applications but poses design challenges. "With SR3 and CDM, we have pushed the performance of diffusion models to the state-of-the-art on super-resolution and class-conditional ImageNet generation benchmarks," the researchers wrote in the blog post.
Windows 11 can be installed on unsupported computers but there’s a big security risk
Bengaluru based Studio Sirah raises $830,000 to develop India-first mobile games — ‘Kurukshetra: Ascension’ will be the first title
Popular on BI
- Silicon Valley needs to stop laying off workers and start firing CEOs
- A laid-off Google engineer describes losing his 'only career' 16 years after starting as an intern
- OpenAI makes a ChatGPT-like tool called Codex that can write software. Here's why Codex won't replace developers and will instead create more demand for their skills.
- Moto e13 with Android 13 (Go Edition) launched in India starting at ₹6,999
- Microsoft’s Bing gets an AI makeover powered by ChatGPT to reset its foot in the battle of new age search engines
- WhatsApp may soon let users schedule calls within group chats
- LinkedIn 100 million members milestone in India—making it the second largest market
- Zoom to lay off around 1,300 employees, CEO Eric Yuan taking 98 per cent pay cut