Please use this identifier to cite or link to this item:
http://hdl.handle.net/1942/42140
Title: | Decision trees and random forests | Authors: | BECKER, Thijs ROUSSEAU, Axel-Jan GEUBBELMANS, Melvin BURZYKOWSKI, Tomasz VALKENBORG, Dirk |
Issue Date: | 2023 | Publisher: | Source: | Neural Processing Letters, 164 (6) , p. 894 -897 | Abstract: | Cost volume is widely used to establish correspondences in optical flow estimation. However, when dealing with low-texture and occluded areas, it is difficult to estimate the cost volume correctly. Therefore, we propose a replacement: feature correlation transformer (FCTR), a transformer with self-and cross-attention alternations for obtaining global receptive fields and positional embedding for establishing correspondences. With global context and posi-tional information, FCTR can produce more accurate correspondences for ambiguous areas. Using position-embedded feature allows the removal of the context network; the positional information can be aggregated within ambiguous motion boundaries, and the number of model parameters can be reduced. To speed up network convergence and strengthen robust-ness, we introduce a smooth L1 loss with exponential weights in the pre-training step. At the time of submission, our method achieves competitive performance with all published optical flow methods on both the KITTI-2015 and MPI-Sintel benchmarks. Moreover, it outperforms all optical flow and scene flow methods in KITTI-2015 foreground-region prediction. | Keywords: | Humans;Decision Trees;Random Forest | Document URI: | http://hdl.handle.net/1942/42140 | ISSN: | 1370-4621 | e-ISSN: | 1573-773X | DOI: | 10.1016/j.ajodo.2023.09.011 | ISI #: | WOS:001125458300001 | Category: | A1 | Type: | Journal Contribution |
Appears in Collections: | Research publications |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
1-s2.0-S0889540623005188-main.pdf Restricted Access | Published version | 326.22 kB | Adobe PDF | View/Open Request a copy |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.