Entrymissing

Medial Feature Superpixels

2021-06-09T20:08:07+02:00

Pixels are the basis of most modern image representations and the starting point of most computer vision algorithms. However, they are actually just artefacts of the image capture and display process. They do not correspond to entities in the real world. Enter superpixels. Superpixels are groupings of pixels into homogeneous regions. They are not yet semantic segmentations of images but reduce the complexity of an image by merging continuous regions into mid-level representations.

During my PhD I worked on some ideas around Superpixel segmentation based on shape-centered features. I’ve just added a summary of the ideas to the publications section. You can also download and read the poster.

Collective Intelligence in Online Groups

2021-06-09T20:08:07+02:00

Collective Intelligence is the general problem solving performance of a team that emerges beyond the abilities of the individuals. Previous studies had shown that this ability is driven by empathy (or more correctly theory-of-mind abilities). But those studies were done with face to face groups. We did a study to see if this also holds true for online groups.

This was a study I did during my PostDoc at MIT. I’ve just added a summary of the ideas to the publications section. You can also download and read the paper.

Detectability of pedestrians

2021-06-04T21:08:07+02:00

You’re likely not going to run over a pedestrian that you’ve noticed. So, the pedestrians that are most in danger are the ones you didn’t see. The ones that blend into the background or are just in unexpected places. A driving assistance system should draw your attention to these kinds of pedestrians to ensure you make the right decisions. During my PhD at the Max Planck Institute I did a study combining experimental psychology and computer vision to predict which pedestrians are hard to notice and model where your attention should be direct to maximize the likelihood that you’re going to notice all pedestrians in a scene.

I’ve just added a summary of the ideas and results of the paper to the publications section. You can also download and read the pdf.

Exploring an inifinite maze in VR

2021-06-02T23:08:07+02:00

One of the more fun studies I did during the beginning of my PhD. We had a tracking hall that allowed free natural movement in VR. It was big but still limited. But, if you introduce small, consistent errors during the movement you can trick the VR users to walk in circles while and make your tracking space functionally infinite.

As it turns out if you walk around a corner in VR and the visual world turns by 90 degrees even though you only turned 80 degrees in the real world you trust your visual system and think you turned the full 90 degrees. If you make the person in VR make enough turns and are smart about introducing these small errors consistently you can trick a person into walking around in circles in the real world while thinking they are walking straight. This means you can walk through infinitely sized virtual worlds in the confines of your moderately sized tracking hall.

I’ve added a summary of the ideas and results of the paper to the publications section. You can also download and read the pdf.

Integrated information and are groups consious?

2021-05-28T20:08:07+02:00

Do groups posess consciousness? It might seem an interesting question from a philosophical point of view but how do we approach this question from a scientific angle?

In a paper I authored together with Tom Malone during my PostDoc at MIT we investigated this idea. The basic premise is that integrated information theory(IIT) offers a framework for calculating a metric that is correlated to the level of consciouness in a system. We applied this metric to in-person teams, wikipedia discussions and internet traffic patterns.

I’ve added a summary of the ideas and results of the paper to the publications section. You can also download and read the pdf.