Shashank Srikanth

In urban driving scenarios, forecasting future trajectories of surrounding vehicles is of paramount importance. While several approaches for the problem have been proposed, the best-performing ones tend to require extremely detailed input representations (eg. image sequences). But, such methods do not generalize to datasets they have not been trained on. We propose intermediate representations that are particularly well-suited for future prediction. As opposed to using texture (color) information, we rely on semantics and train an autoregressive model to accurately predict future trajectories of traffic participants (vehicles) (see fig. above). We demonstrate that using semantics provides a significant boost over techniques that operate over raw pixel intensities/disparities. Uncharacteristic of state-of-the-art approaches, our representations and models generalize to completely different datasets, collected across several cities, and also across countries where people drive on opposite sides of the road (left-handed vs right-handed driving). Additionally, we demonstrate an application of our approach in multi-object tracking (data association). To foster further research in transferrable representations and ensure reproducibility, we release all our code and data.

In 2015, 391,000 people were injured due to distracted driving in the US. One of the major reasons behind distracted driving is the use of cell-phones, accounting for 14% of fatal crashes. Social media applications have enabled users to stay connected, however, the use of such applications while driving could have serious repercussions - often leading the user to be distracted from the road and ending up in an accident. In the context of impression management, it has been discovered that individuals often take a risk (such as teens smoking cigarettes, indulging in narcotics, and participating in unsafe sex) to improve their social standing. Therefore, viewing the phenomena of posting distracted driving posts under the lens of self-presentation, it can be hypothesized that users often indulge in risk-taking behavior on social media to improve their impression among their peers. In this paper, we first try to understand the severity of such social-media-based distractions by analyzing the content posted on a popular social media site where the user is driving and is also simultaneously creating content. To this end, we build a deep learning classifier to identify publicly posted content on social media that involves the user driving. Furthermore, a framework proposed to understand factors behind voluntary risk-taking activity observes that younger individuals are more willing to perform such activities, and men (as opposed to women) are more inclined to take risks. Grounding our observations in this framework, we test these hypotheses on 173 cities across the world. We conduct spatial and temporal analysis on a city-level and understand how distracted driving content posting behavior changes due to varied demographics. We discover that the factors put forth by the framework are significant in estimating the extent of such behavior.