Eco3D is an initiative started by Jeppe Revall Frisvad, Knut Conradsen, and Henrik Aanæs to coordinate our research activities in sensors and 3D geometric and radiometric modeling in relation to the cyber-physical 3D ecosystem. This ecosystem is an abstraction of the fact that we oftentimes wish to manipulate the physical world based on sensor generated digital or cyber models of this physical world. This is illustrated in the following diagram.
This ecosystem also encompasses much of the research we do at the Section for Image Analysis and Computer Graphics at the Technical University of Denmark, and is as such a nice framework for exploiting the possible synergies within our group. Examples of this ecosystem are
As seen under People, this initiative has a sufficient number of participants, working on parts of the ecosystem. A benefit of taking the holistic ecosystem approach to our research is that it also enforces relevant performance metrics on our research in the individual parts. In 2015, the first year of the initiative, our collaboration has focus on
We investigate the process of optical surface scanning to see if the statistical quality can reach a level high enough for such a scanner to be used as a full grade metrological device.
We focus on digitalization and quality control of the 3D printing process. We also develop and evaluate novel 3D printing (additive manufacturing) systems for use within the design space of manufacturing and rapid prototyping.
We create efficient rendering techniques and mathematical models that predict the appearance of materials from their physical microstructure. The goal is to interactively compute realistic images of real-world materials.
Having densly sampled BRDFs of the surface of each object in a given scene allows for photorealistic rendering of that scene. This project aims at better acquisition of surface BRDFs in 3D scenes composed of simple geometries.
We create a realistic dataset for evaluation of Non-Rigid Structured From Motion algorithms. The dataset provides a dense point set with an accurate ground truth. This is accomplished through stop-motion animated animatronics with dense structured light scanning for each frame.
We have created a multi-view stereo (MVS) dataset where the focus is on reconstruction of geometry and appearance. Three glass objects with different radiometric properties are selected and multiple MVS scenes are captured, reconstructed and rendered.