About me

I am a PhD student in Computer Vision at Faculty of Electrical Engineering and Computing, University of Zagreb. My advisor and life coach is Siniša Šegvić.

My current research focuses on future anticipation in driving scenarios. The ability of machines to anticipate and reason about the future could be crucial in development and integration of autonomous driving systems. It allows planning of actions not only by looking at the past, but at the forecasted future as well. For example, knowing future location of pedestrian could enable potentially life-saving decisions. Currently, I am more more oriented towards the forecasting at the semantic level (e.g. semantic segmentation), rather than raw RGB frames. However, raw frame forecasting is also interesting and challenging problem.

In my free time I like to hang out with my friends, ride a bike and attend a pub quiz with my fantastic team DanceNet-121. Now, I hope to publish some interesting posts on my blog about computer vision and stuff.

Recent work

  • [arxiv] Multi-domain semantic segmentation with pyramidal fusion.
    Marin Oršić, Petra Bevandić, Ivan Grubišić, Josip Šarić, Siniša Šegvić.
    Tech report for RVC20.
  • [cvf] Warp to the Future: Joint Forecasting of Features and Feature Motion.
    Josip Šarić, Marin Oršić, Tonći Antunović, Sacha Vražić, Siniša Šegvić.
    CVPR 2020.
  • [arxiv] Single Level Feature-to-Feature Forecasting with Deformable Convolutions.
    Josip Šarić, Marin Oršić, Tonći Antunović, Sacha Vražić, Siniša Šegvić.
    GCPR 2019.
  • Estimating prediction uncertainty of discriminative convolutional models with adversarial learning. J. Šarić. Master’s thesis. University of Zagreb, Faculty of Electrical Engineering and Computing. Zagreb, 2018.