AI2FUTURE2019 / 2 days / 4 keynotes / 20+ moving presentation

10th & 11th of October 2019, Kraš Auditorium, Zagreb, Croatia
Associate professor, Department of Electronics, Microelectronics, Computer and Intelligent Systems

Siniša Šegvić

Associate professor, Department of Electronics, Microelectronics, Computer and Intelligent Systems

Siniša Šegvić completed his PhD degree in 2004 at the University of Zagreb. He spent one year as a postdoc researcher at IRISA Rennes (2006) and another year as a postdoc researcher at TU Graz (2007). Currently he is an associate professor at the University of Zagreb. He is a program committee member of the VISAPP 2018 conference and an associate editor of the Journal of Computing and Information Technology. His research expertise is in the fields of computer vision and deep learning, with special interest in applications for autonomous vehicles and safe traffic. He published more than 50 national and international scientific papers.



Lightweight convolutional models for real-time dense prediction and forecasting

Convolutional models have become an indispensable ingredient for automated visual recognition. Most of the existing work in the field is based on heavyweight models with hundreds of layers. Unfortunately, such designs are unable to offer real-time performance on embedded hardware, which precludes many important applications such as autonomous driving. However, recent work shows that convolutional models are extremely resistant to overfitting, contrary to established theories of machine learning.

This suggests that useful performance could be obtained with much less computational effort. Our latest findings support this notion in the fields of semantic segmentation prediction and forecasting. In particular, we demonstrate that the model capacity can be effectively compensated by careful wiring of information data paths within the model. Our approaches achieve state-of-the-art ratio between recognition accuracy and processing time,  and show that real-time applications of convolutional models are indeed feasible.



Follow Siniša





There are many sessions and workshops at the conference, please select the ones that are great for you, but please also reserve some time for networking and meeting other great people.


Deep Learning for Smart Vehicles and Safe Roads

Recent advances in deep convolutional models have caused unprecedented growth of computer vision performance. This has opened exciting applications in the fields of smart vehicles and safe roads. Pixel-level image understanding can be achieved by associating each image window with a meaningful class such as ‘road’, ‘terrain’, ‘sidewalk’ or ‘person’. The resulting semantic map reveals the kind of surface terrain in front of the vehicle, and may be used to recover the traversability map required for motion planning. Depth can be recovered by predicting a disparity field which maximizes similarity between two stereo images.
The resulting dense reconstruction can be used to detect obstacles and potholes. Classification of short video clips allows simultaneous detection of a variety of road safety attributes in video. We illustrate these opportunities by experiments on large natural images from public datasets Cityscapes and KITTI, as well as on a large video dataset collected by an industrial partner. The obtained results confirm feasibility of the presented approaches on commodity and embedded hardware.

Read more