IC Student Hannes Hergeth's last terrarium wins rendering competition

© 2018 EPFL / Hannes Hergeth

© 2018 EPFL / Hannes Hergeth

Realistic rendering is challenging, as it involves physics; the properties of light and materials, applied mathematics; solving the equations of light transport, and computer science; numerical computation and data structures.

Students of Professor Wenzel Jakob’s (Realistic Graphics Lab) course on the subject learn advanced 3D graphics techniques to produce realistic visual renderings and to understand how light behaves and interacts with objects and materials such as wood, metal, and glass. The objective is to recreate these phenomena in a computer simulation to develop synthetic images indistinguishable from photographs.

At the end of the semester, the students realize a final project where they compete for the best image rendering, judged by world-renowned experts from research and the visual effects industry. This year the expert panel, consisting of André Mazzone (Industrial Light and Magic), Laurent Belcour (Unity Labs) and Marios Papas (Disney Research), chose Hannes Hergeth’s image as the best rendering on the theme of “The Last One.”

What inspired you and what was your aim?

I featured terrariums because I like how they encapsulate small-scale worlds while providing a nice visual balance between organic and synthetic components. The variation in real-world terrariums also allowed me to tailor the scene to highlight specific materials with a chosen geometry.

To create the image of “The Last One,” I built a scene as if the terrarium construction was underway. This included a messy workspace and a single plant — the “Last One” — left outside of the terrariums, still waiting to be planted. The theme led me to a more compelling image by forcing me to attach a story to my terrariums, and to think about the details to include in the image to tell that story, such as a shovel, loose soil, etc. Being reviewed by people helped me improve my story, as my Dad suggested the addition of the newspapers as well as a more natural arrangement of the objects themselves.

What was particularly challenging in the rendering process?

I invested a lot of effort into the materials for my scene, incorporating realistic materials like brushed and scratched metals. While visually interesting, these were challenging to simulate convincingly and my approach, therefore, required creativity.

The way in which light interacts with plants is also very complex, and I did not have enough time to implement a physically correct model. As humans are sensitive to inconsistencies in natural things, to preserve realism, I was careful to select only the plants, which my slightly naive technique could handle.

What would your recommendations be for students attempting this task in the future?

For me, it was very helpful to collect many inspirational images and note what I liked about each one. These served as good reference points and validation for my final image. I would also avoid tackling too big of a scene. If you focus on something specific and manageable, you can make it great by investing time in realistic details.

Lastly, do not underestimate the influence of the scene’s non-rendering elements: geometry, textures, lighting, and overall composition. While good renderers can produce a nice realization of their input, renderers cannot produce great images from arbitrary inputs.