Matches in SemOpenAlex for { <https://semopenalex.org/work/W3107163670> ?p ?o ?g. }
- W3107163670 endingPage "2411" @default.
- W3107163670 startingPage "2400" @default.
- W3107163670 abstract "SegBlocks reduces the computational cost of existing neural networks, by dynamically adjusting the processing resolution of image regions based on their complexity. Our method splits an image into blocks and downsamples blocks of low complexity, reducing the number of operations and memory consumption. A lightweight policy network, selecting the complex regions, is trained using reinforcement learning. In addition, we introduce several modules implemented in CUDA to process images in blocks. Most important, our novel BlockPad module prevents the feature discontinuities at block borders of which existing methods suffer, while keeping memory consumption under control. Our experiments on Cityscapes, Camvid and Mapillary Vistas datasets for semantic segmentation show that dynamically processing images offers a better accuracy versus complexity trade-off compared to static baselines of similar complexity. For instance, our method reduces the number of floating-point operations of SwiftNet-RN18 by 60% and increases the inference speed by 50%, with only 0.3% decrease in mIoU accuracy on Cityscapes." @default.
- W3107163670 created "2020-12-07" @default.
- W3107163670 creator A5056593702 @default.
- W3107163670 creator A5074816094 @default.
- W3107163670 date "2023-02-01" @default.
- W3107163670 modified "2023-09-30" @default.
- W3107163670 title "SegBlocks: Block-Based Dynamic Resolution Networks for Real-Time Segmentation" @default.
- W3107163670 cites W1613249581 @default.
- W3107163670 cites W1996901117 @default.
- W3107163670 cites W2058641082 @default.
- W3107163670 cites W2108598243 @default.
- W3107163670 cites W2109255472 @default.
- W3107163670 cites W2171943915 @default.
- W3107163670 cites W2194775991 @default.
- W3107163670 cites W2233116163 @default.
- W3107163670 cites W2340897893 @default.
- W3107163670 cites W2412782625 @default.
- W3107163670 cites W2560023338 @default.
- W3107163670 cites W2562731582 @default.
- W3107163670 cites W2604319603 @default.
- W3107163670 cites W2606492274 @default.
- W3107163670 cites W2609402060 @default.
- W3107163670 cites W2762439315 @default.
- W3107163670 cites W2781228439 @default.
- W3107163670 cites W2804199516 @default.
- W3107163670 cites W2884751099 @default.
- W3107163670 cites W2889469641 @default.
- W3107163670 cites W2956252641 @default.
- W3107163670 cites W2962707357 @default.
- W3107163670 cites W2962772649 @default.
- W3107163670 cites W2962944050 @default.
- W3107163670 cites W2963091558 @default.
- W3107163670 cites W2963163009 @default.
- W3107163670 cites W2963419596 @default.
- W3107163670 cites W2963881378 @default.
- W3107163670 cites W2963896595 @default.
- W3107163670 cites W2964217532 @default.
- W3107163670 cites W2981669902 @default.
- W3107163670 cites W2981689412 @default.
- W3107163670 cites W2981899103 @default.
- W3107163670 cites W2984005812 @default.
- W3107163670 cites W2997806675 @default.
- W3107163670 cites W2998317038 @default.
- W3107163670 cites W3035678286 @default.
- W3107163670 cites W3082097505 @default.
- W3107163670 cites W3108450508 @default.
- W3107163670 cites W3109632933 @default.
- W3107163670 cites W3110440461 @default.
- W3107163670 cites W3128896552 @default.
- W3107163670 cites W3172265425 @default.
- W3107163670 cites W3194677016 @default.
- W3107163670 cites W3196904463 @default.
- W3107163670 cites W4214717370 @default.
- W3107163670 cites W4241811150 @default.
- W3107163670 doi "https://doi.org/10.1109/tpami.2022.3162528" @default.
- W3107163670 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/35349431" @default.
- W3107163670 hasPublicationYear "2023" @default.
- W3107163670 type Work @default.
- W3107163670 sameAs 3107163670 @default.
- W3107163670 citedByCount "4" @default.
- W3107163670 countsByYear W31071636702023 @default.
- W3107163670 crossrefType "journal-article" @default.
- W3107163670 hasAuthorship W3107163670A5056593702 @default.
- W3107163670 hasAuthorship W3107163670A5074816094 @default.
- W3107163670 hasBestOaLocation W31071636702 @default.
- W3107163670 hasConcept C111919701 @default.
- W3107163670 hasConcept C11413529 @default.
- W3107163670 hasConcept C115961682 @default.
- W3107163670 hasConcept C124504099 @default.
- W3107163670 hasConcept C134306372 @default.
- W3107163670 hasConcept C138885662 @default.
- W3107163670 hasConcept C153180895 @default.
- W3107163670 hasConcept C154945302 @default.
- W3107163670 hasConcept C15627037 @default.
- W3107163670 hasConcept C179799912 @default.
- W3107163670 hasConcept C2524010 @default.
- W3107163670 hasConcept C2776214188 @default.
- W3107163670 hasConcept C2776401178 @default.
- W3107163670 hasConcept C2777210771 @default.
- W3107163670 hasConcept C31972630 @default.
- W3107163670 hasConcept C33923547 @default.
- W3107163670 hasConcept C41008148 @default.
- W3107163670 hasConcept C41895202 @default.
- W3107163670 hasConcept C50644808 @default.
- W3107163670 hasConcept C89600930 @default.
- W3107163670 hasConcept C98045186 @default.
- W3107163670 hasConceptScore W3107163670C111919701 @default.
- W3107163670 hasConceptScore W3107163670C11413529 @default.
- W3107163670 hasConceptScore W3107163670C115961682 @default.
- W3107163670 hasConceptScore W3107163670C124504099 @default.
- W3107163670 hasConceptScore W3107163670C134306372 @default.
- W3107163670 hasConceptScore W3107163670C138885662 @default.
- W3107163670 hasConceptScore W3107163670C153180895 @default.
- W3107163670 hasConceptScore W3107163670C154945302 @default.
- W3107163670 hasConceptScore W3107163670C15627037 @default.
- W3107163670 hasConceptScore W3107163670C179799912 @default.
- W3107163670 hasConceptScore W3107163670C2524010 @default.
- W3107163670 hasConceptScore W3107163670C2776214188 @default.