Whenever a new product is launched in the market, the first thing that catches the consumer’s eye is the look. The movie, ‘Avatar’ was launched to rave reviews that praised its CGI. As much as we may take graphics for granted, designers put in painstaking efforts to create visually stunning products. It is not as easy as it may seem. Fortunately, the job has been made easier by the advancements in Artificial Intelligence (AI). AI is the one industry that is revolutionizing every other industry in its wake. Let’s look into how AI has helped improve our overall visual experience.
A few years ago, it would have been hard to imagine that the field of graphics could be automated using AI. Today, we’ve come to a point where tools like Adobe’s Sensei help create an eye-catching experience through Machine Learning. It works as your assistant and helps design the right content by suggesting tips that can bring life into the content. It can identify objects in photos and organize them so that you can search and select the right pictures for your job. It seamlessly integrates with other Adobe tools to make the workflow convenient and efficient.
When graphics designers make something, they use polygons (usually triangles) and rasterization to create 3D images.
Lots of these polygons are joined together piece by piece and rendered to give realistic images. This approach made programming an environment easier but the trade-off was that the physics of a given scene was static. Ray Tracing has changed that.
Ray Tracing (RT) and Deep Learning Super Sampling (DLSS) are technologies NVIDIA introduced in their RTX 2000 series GPUs. Ray Tracing is a rendering technique where images are rendered by calculating how light rays interact with objects in real time.
While RT has existed for some time now, it is a slow process. Traditionally, it required many Graphics Processing Units (GPUs) to work together in parallel, making the process computationally expensive and complicated to program. With advancements in AI, the whole process is computationally cheaper than before.
AI accelerated Ray Tracing leverages the power of AI to calculate light interactions efficiently. It also helps developers avoid headaches during production as they can use the GPU’s capability to calculate the light interaction instead of programming the interaction themselves.
Today, many programs offer support for RT. Games like Cyberpunk 2077 use it to provide pleasing visual experiences, and productivity software like Blender uses it to accelerate work and production through fast rendering.
Deep Learning Super Sampling (DLSS)
While Ray Tracing brings some exciting improvements to how we experience graphics, it is an expensive process even with AI acceleration. It is not uncommon to see programs slow down as a result of enabling RT. This is where DLSS comes into the picture.
Conventionally, images are rendered at the resolution specified. Higher resolutions are computationally expensive to render and store in the memory and then display. DLSS removes this bottleneck.
Deep Learning Super Sampling (DLSS) is an upscaling technique where instead of rendering an image natively, it is rendered at a lower resolution, and then, AI is used to patch up the image to a higher resolution. For example, to render a 2K image, DLSS enables you to render it at a lower resolution (say 1080p) and then scale it up to a 2K resolution. This helps reduce the GPU’s workload by helping it output frames at a faster rate. These frames are then patched up by an AI to give the final output. Images rendered this way are indistinguishable from natively rendered images. In some cases, they look even sharper than the natively rendered images!
DLSS has made Ray Tracing significantly cheaper, thus keeping the performance at acceptable levels and at the same time, helping users experience the power of Ray Tracing.
We explored some exciting new features in the world of graphics made possible by AI. Improvements in AI have helped accelerate the workflow and improved the output of this industry at a rate never seen before.