Equations are not being displayed properly on some articles. We hope to have this fixed soon. Our apologies.

Sparavigna, A. (2017). 3D Faces from 2D Pictures. PHILICA.COM Article number 1121.

ISSN 1751-3030  
Log in  
Register  
  1203 Articles and Observations available | Content last updated 19 October, 04:02  
Philica entries accessed 3 365 465 times  


NEWS: The SOAP Project, in collaboration with CERN, are conducting a survey on open-access publishing. Please take a moment to give them your views

Submit an Article or Observation

We aim to suit all browsers, but recommend Firefox particularly:

3D Faces from 2D Pictures

Amelia Carolina Sparavignaunconfirmed user (Department of Applied Science and Technology, Politecnico di Torino)

Published in compu.philica.com

Abstract
A very difficult problem to solve is that of having 3D facial reconstructions from single 2D images. Here we discuss a remarkable software solving this task, developed by Jackson, Bulat, Argyriou and Tzimiropoulos at the University of Nottingham and Kingston University. The researchers have obtained their result by training a Convolutional Neural Network (CNN) on an appropriate dataset consisting of 2D images and 3D facial models.

Article body

 

3D Faces from 2D Pictures

 

Amelia Carolina Sparavigna

Politecnico di Torino

 

A very difficult problem to solve is that of having 3D facial reconstructions from single 2D images. Here we discuss a remarkable software solving this task, developed by Jackson, Bulat, Argyriou and Tzimiropoulos at the University of Nottingham and Kingston University. The researchers have obtained their result by training a Convolutional Neural Network (CNN) on an appropriate dataset consisting of 2D images and 3D facial models.

 

The reconstruction of a 3D surface from a 2D map is a problem that can be of general interest for several applications. For instance, if we have a 2D image from a scanning electron microscope, we can render it in a 3D surface as discussed in [1,2]. A simple approach is that of assuming the grey tones of the pixels of images as corresponding to the z-coordinates of the points at x,y coordinates [3]. This approach is quite natural for images obtained from the scan of a surface.

Problems exist which are far more difficult to solve, concerning a 3D rendering from a 2D map. One of them is the 3D reconstruction of a face from a single 2D image. Actually, a remarkable solution exists and it is given in [4]. The researchers, Aaron S. Jackson, Adrian Bulat, Vasileios Argyriou and Georgios Tzimiropoulos, at the University of Nottingham and Kingston University, have developed a software, which is answering this specific task, by training a Convolutional Neural Network (CNN) on an appropriate dataset consisting of 2D images and 3D facial models. The researchers are also providing a site for experiments, at the address http://www.cs.nott.ac.uk/~psxasj/3dme/index.php . Let us call here this software as “3DME”. For experiments, a frontal image is required. After the image is uploaded, the site brings us to the corresponding 3D model, which is easy to move by means of the mouse. An OBJ file of the 3D model is also given.

After some experiments, we can easily see that any picture is good for a 3D rendering; that is, we can use also a painted portray of a person to have a 3D face. An example is shown in the Figure 1. We see Benjamin Franklin in a portrait by Joseph Duplessis ca. 1785. We have, in the upper/left panel, the original portrait; in the upper/right panel the output of software 3DME at the above-mentioned site. In the lower images, we see three different positions of the face that we can have from software. The result is excellent. Of course, we can easily imagine that this software can be used to create digital 3D portraits of historical figures.

 

 

Figure 1: In the upper/left panel, we see Benjamin Franklin in a portrait by Joseph Duplessis ca. 1785. In the upper/right panel, it is shown the output of software. In the lower images, we see three different positions of the face that we can have obtain by rotating the model with the mouse.

 

In some previous papers [5-7], we have investigated a software of facial transformation (in20years.com), applying it to the pictures of marble busts. Using in20years.com software, the pictures of marble busts turn into pictures that look like those of real persons. Therefore, let us apply software 3DME on a marble bust too. That is, let us investigate 3DME software on a picture of a marble bust to see if this software is able to give a reliable 3D model of the original bust. Let us use the marble bust found in the Rhone River near Arles. Some scholars consider it a portrait of Julius Caesar. Thanks to Butko, Wikipedia, we have two images of this bust, here shown in the Figure 2. In the Figure 3, we give the results obtained by 3DME. Comparing the images, we can conclude that the result of the method proposed in [4] is excellent.

 

Figure 2: The marble bust found in the Rhone River near Arles (Courtesy Butko, Wikipedia).

 

Figure 3: Results of the 3DME software on the left image of Figure 2. In the middle, the 3D model is given without background.

 

References

[1] Groeber, M. A., Haley, B. K., Uchic, M. D., Dimiduk, D. M., & Ghosh, S. (2006). 3D reconstruction and characterization of polycrystalline microstructures using a FIB–SEM system. Materials Characterization, 57(4), 259-273.

[2] Tafti, A. P., Kirkpatrick, A. B., Alavi, Z., Owen, H. A., & Yu, Z. (2015). Recent advances in 3D SEM surface reconstruction. Micron, 78, 54-66.

[3] Sparavigna, A. C., Giorcelli, M., & Guastella, S. (2017). Three-dimensional rendering of Biochar turfaces from Their FESEM images. Biochar: Production, Characterization and Applications, Aug 2017, Alba, Italy. 2017, <http://www.engconf.org/biochar-production-characterization-and-applications/>. <hal-01577075>

[4] Jackson, A. S., Bulat, A., Argyriou, V., & Tzimiropoulos, G. (2017). Large pose 3D face reconstruction from a single image via direct volumetric CNN regression. arXiv preprint arXiv:1703.07834.

[5] Sparavigna, A. C. (2013). A software for aging faces applied to ancient marble busts. arXiv preprint arXiv:1304.1022.

[6] Sparavigna, A. C. (2016). The appearance of roman emperors rendered by a face detection software (April 30, 2016). Available at SSRN: https://ssrn.com/abstract=2773502 or http://dx.doi.org/10.2139/ssrn.2773502

[7] Sparavigna, A. C. (2013). Facial transformations of ancient portraits: the face of Caesar. arXiv preprint arXiv:1304.1972.



Information about this Article
This Article has not yet been peer-reviewed
This Article was published on 22nd September, 2017 at 19:06:10 and has been viewed 351 times.

Creative Commons License
This work is licensed under a Creative Commons Attribution 2.5 License.
The full citation for this Article is:
Sparavigna, A. (2017). 3D Faces from 2D Pictures. PHILICA.COM Article number 1121.


<< Go back Review this ArticlePrinter-friendlyReport this Article



Website copyright © 2006-07 Philica; authors retain the rights to their work under this Creative Commons License and reviews are copyleft under the GNU free documentation license.
Using this site indicates acceptance of our Terms and Conditions.

This page was generated in 0.2905 seconds.