1: \begin{abstract}

2:     We present a fast and simple technique to convert images into an emissive

3:     surface-based scene representation. Building on existing emissive volume

4:     reconstruction algorithms, we introduce a subtle yet impactful modification

5:     of the loss function requiring changes to only a few lines of code:

6:     instead of integrating the radiance field along rays and supervising the

7:     resulting images, we project the training images into the scene to directly

8:     supervise the spatio-directional radiance field.

9: 

10:     The primary outcome of this change is the complete removal of alpha

11:     blending and ray marching from the image formation model, instead moving

12:     these steps into the loss computation. In addition to promoting

13:     convergence to surfaces, this formulation assigns explicit semantic meaning

14:     to 2D subsets of the radiance field, turning them into well-defined

15:     emissive surfaces. We finally extract a level set from this representation,

16:     which results in a high-quality emissive surface model.

17: 

18:     Our method retains much of the speed and quality of the baseline algorithm.

19:     For instance, a suitably modified variant of Instant~NGP maintains

20:     comparable computational efficiency, while achieving an average PSNR that

21:     is only 0.1 dB lower. Most importantly, our method generates explicit

22:     surfaces in place of an exponential volume, doing so with a level of

23:     simplicity not seen in prior work.

24: \end{abstract}

25: