cudaIFT: 180x Faster Image Foresting Transform for Waterpixel Estimation using CUDA

Gonçalves, Henrique M.; Vasconcelos, Gustavo J. Q.; Rangel, P.; Carvalho, Murilo; Archilha, Nathaly L.; Spina, Thiago Vallin

doi:10.5220/0007402703950404

Cited by 4 publications

(2 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, when comparing Ω = 5 and Ω = 10 in all datasets, there are no significant improvements, indicating a possible upper bound (i.e., performing unnecessary iterations). Therefore, we limited SICLE to perform, at most, five iterations using the ROOT function, which allows future improvements [16,13].…”

Section: Ablation Studymentioning

confidence: 99%

Efficient Multiscale Object-based Superpixel Framework

Belém¹,

Cousty²,

Guimarães³

et al. 2022

Preprint

View full text Add to dashboard Cite

Superpixel segmentation can be used as an intermediary step in many applications, often to improve object delineation and reduce computer workload. However, classical methods do not incorporate information about the desired object. Deep-learning-based approaches consider object information, but their delineation performance depends on data annotation. Additionally, the computational time of object-based methods is usually much higher than desired. In this work, we propose a novel superpixel framework, named Superpixels through Iterative CLEarcutting (SICLE), which exploits object information being able to generate a multiscale segmentation onthe-fly. SICLE starts off from seed oversampling and repeats optimal connectivity-based superpixel delineation and object-based seed removal until a desired number of superpixels is reached. It generalizes recent superpixel methods, surpassing them and other state-of-the-art approaches in efficiency and effectiveness according to multiple delineation metrics.

show abstract

Section: Ablation Studymentioning

confidence: 99%

Efficient Multiscale Object-based Superpixel Framework

Belém¹,

Cousty²,

Guimarães³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Finally, the speed-up of ISF2SVX over GB is 0.95, being slightly slower than GB. However, due to recent findings [10,6] and for a suitable definition of components (e.g, GRID sampling and w 3 arc-cost function), it is possible to further improve the speed of ISF2SVX without prejudicing the object delineation performance.…”

Section: Qualitative Analysismentioning

confidence: 99%

Graph-Based Supervoxel Computation from Iterative Spanning Forest

Almeida

Belém

Carneiro

et al. 2021

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Supervoxel segmentation leads to major improvements in video analysis since it generates simpler but meaningful primitives (i.e., supervoxels). Thanks to the flexibility of the Iterative Spanning Forest (ISF) framework and recent strategies introduced by the Dynamic Iterative Spanning Forest (DISF) for superpixel computation, we propose a new graph-based method for supervoxel generation by using iterative spanning forest framework, so-called ISF2SVX, based on a pipeline composed by four stages: (a) graph creation; (b) seed oversampling; (c) IFT-based superpixel delineation; and (d) seed set reduction. Moreover, experimental results show that ISF2SVX is capable of effectively describing the video's color variation through its supervoxels, while being competitive for the remaining metrics considered.

show abstract