Ground Truth Via Games
Road accidents are a major thread to societyAccording to the World Health Organization in Europe, every year 127 thousand people are killed and around 2.4 million are injured every year by traffic accidents. If you imagine, the population of a medium-sized city such as Ingolstadt is wiped out every year. According to a public consultation of the European Union in 2012, road injuries are the second-largest thread to society, directly after unemployment.
How to reduce road accidents?A human usually reacts within about a second if something unexpected happens. What does this mean? Let's consider an urban situation. A person drives with about 50km/h. Suddenly, a ball jumps onto the street. The driver will drive around 13 more meters before he will react. Once the driver understood the threat, he usually doesn' break strong enough, it might take another 13-30 meters until the car comes to a halt.
Driver assistance systems, such as automatic emergency breaks, are a key factor for reducing such accidents. The system will react in milliseconds and then break with full force if necessary. Instead of stopping after about 30 meters, the car will always come to a halt after just about 15 meters.
What we do to achieve thisSuch driver assistance systems often use cameras to record what the driver sees. These images are analyzed for threads the driver might not have noticed. The Heidelberg Collaboratory for Image Processing (HCI) is involved in understanding how well such systems work.
To achieve this, we recorded a lot of videos showing many different situations a driver assistance system should understand. This was possible with the help of many actors, who stepped into the role of a pedestrian.
We need your help!But for now, no automated algorithm is capable of providing the data in the needed level of quality. Yet, humans can detect features in images that remain hidden for the computer. Unfortunately, we have thousands and thousands of images and there are not enough people in our offices to analyze them. This is why we ask you for help!
The HCI and Bigpoint are cooperating on a project for annotation of our traffic videos. Therefore, we need you to find people and cars in the scene, outline them and give us some more hints on how to understand the motion within the video.
With the data collected during this research project, we can evaluate the quality of driver assistance systems. Therefore, with your help, we can finally save thousands of lives!
During February 2014 we want to see how you like our approach and how many of you are willing to help us. Your feedback is important to us, so please use the comment function in our system as often as you seem fit.
We will keep you posted with information on how we proceed with your contribution as the project evolves. Thanks for your help! You are awesome!