Towards more efficient stereo matching with CNNs

Add to your list(s) Download to your calendar using vCal

• Giorgos Zampokas
• Monday 08 March 2021, 16:00-17:00
• Teams (in the "CAS Seminars" team).

If you have a question about this talk, please contact George A Constantinides.

Stereo matching is another computer vision task, which is currently dominated by recent CNN methods, achieving state-of-the-art performance. However, their improved accuracy comes with increased computational complexity, which remains an issue for real-time applications, especially in less powerful hardware. It is important to come up with more flexible implementations, which can adapt to different hardware and requirements, introducing a trade-off between performance and computational complexity.

First, this talk introduces how CNNs solve the problem of stereo matching and outline the critical components in terms of computational budget. Our current work is motivated by the redundancy of computations in some image regions, which result in minimal impact in the final result when processed. Thus, we work on a data-driven methodology to guide the network to assign expensive computations to more important regions, whereas other image areas are processed with computationally cheaper alternatives.

This talk is part of the CAS Talks series.

This talk is included in these lists:

• Andrea Picciau's list
• CAS Talks
• Circuits and Systems Group: Internal Seminars
• Teams (in the "CAS Seminars" team)

Note that ex-directory lists are not shown.