We are offering an internship position for a “3D Artist for Generating Realistic Textures for Everyday 3D Object Meshes”. The project will take about 120-160h, depending on your qualifications.
The Perceiving Systems Department is offering an internship position for a “3D Artist for Generating Realistic Textures for Everyday 3D Object Meshes”. A general goal of our bigger research agenda is to capture, understand, and analyze people interacting with their surrounding everyday objects. A recently developed method of ours accurately captures real human subjects interacting with everyday 3D printed objects, and reconstructs them as 3D meshes that interact over time. Overall we captured 10 subjects interacting with ~50 objects.
Figure 1: Example frames from our dataset. Please note the lack of realistic appearance; humans are shown with blue and objects with yellow color. The examples show (left to right): eating a banana, screwing a light bulb, handing over binoculars, taking a photo.
Figure 1 shows some example frames from our dataset. Please note that the dataset in its current form contains only geometry meshes without natural appearance; in Figure 1 humans are shown with blue color and objects with yellow color. Upon release of the dataset, we want to include realistic appearance textures for both the human and the object so that, in contrast to Figure 1, the human and the object look as realistic as possible. For each human subject and object mesh it is important to have a variety of textures - this is important for rendering synthetic images as varied training data for Machine Learning applications.
For humans, our group has internal tools to automate the creation of human texture maps. For this, we register our statistical 3D human model SMPL to 3D human scans (captured by us or purchased by artists) and “transfer” colors from the scans to SMPL, using also the SMPL UV map.
For objects, we have no such automated tools. The internship holder will therefore learn how to work on creating realistic texture maps for our objects. Figure 2 shows examples of 3D objects without (left) and with (right) realistic texture, while Figures 3 shows an intuitive example for texturing with a UV map; texturing a 3D mesh is like wrapping a colorful aluminum foil on a 3D chocolate.
Figure 2: Example 3D meshes of everyday objects. For each pair we show a visualization of (left) only geometry and (right) geometry with realistic texture/appearance.
Figure 3: Intuitive example for texturing a 3D mesh. Think of the 3D mesh as a chocolate santa claus (brown color), the 2D texture image as the unfolded colorful aluminium foil, and the UV map as the set of correspondences between the two. Texturing is like wrapping the aluminum foil (texture) around the chocolate (mesh), so that (UV map) the eyes/nose/etc of the foil fall onto the chocolate eyes/nose/etc.
Figure 4: (Left) A 3D mesh of a hammer without any texture, i.e. geometry-only visualization. (Right) Applying a “wood” 2D texture image on the handle of the 3D hammer mesh.
Figure 5: Examples of various material and textures for the head of a 3D hammer. Rich appearance variance is important for using such synthetic images as training data for Machine Learning applications.
Our end goal is to use our meshes and textures for rendering synthetic images that can be used to train Machine Learning models. For this, rich appearance variation is crucial. Therefore we need several appearances per object as seen in Figure 5 for the hammer head, albeit for all object parts. We would like to use a single UV map per object, so that we release to the public only a single 3D mesh file per object that can automatically “load” different textures.
The internship holder needs some prior knowledge with industry-standard 3D graphics and texturing tools (e.g. Blender, Autodesk Maya, Adobe Substance3D) and an artistic taste on creating realistic object appearances. The necessary software licenses will be provided by MPI for the duration of the project.
We estimate the internship to take 120-160 hours, depending on prior experience. For a part-time position of 10h/week or 40h/month (25% position), we expect a 3-4 month internship to cover the project needs. We expect the following indicative timeline for the internship holder:
1st month: The intern (1) understands the project needs and goals, (2) understands the structure and use of our data, (3) establishes a “3D modeling pipeline” with commercial software to load and process our data, (4) iterates on a few examples and takes feedback from researchers to refine the “3D modeling pipeline” and also refine the deliverable goals.
The position is available as of now. Early starting dates are strongly preferred. The position is open until filled.
Please direct your applications to firstname.lastname@example.org and mention "Internship t PS" in the subject line.
The Max Planck Society is committed to increasing the number of individuals with disabilities in its workforce and therefore encourages applications from such qualified individuals. The Max Planck Society strives for gender equality and diversity. Furthermore, the Max Planck Society seeks to increase the number of women in its workforce in those areas where they are underrepresented and therefore explicitly encourages women to apply.