Matches in SemOpenAlex for { <https://semopenalex.org/work/W3163130003> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W3163130003 abstract "Deep convolutional neural networks have achieved remarkable progress in recent years. However, the large volume of intermediate results generated during inference poses a significant challenge to the accelerator design for resource-constraint FPGA. Due to the limited on-chip storage, partial results of intermediate layers are frequently transferred back and forth between on-chip memory and off-chip DRAM, leading to a non-negligible increase in latency and energy consumption. In this paper, we propose block convolution, a hardware-friendly, simple, yet efficient convolution operation that can completely avoid the off-chip transfer of intermediate feature maps at run-time. The fundamental idea of block convolution is to eliminate the dependency of feature map tiles in the spatial dimension when spatial tiling is used, which is realized by splitting a feature map into independent blocks so that convolution can be performed separately on individual blocks. We conduct extensive experiments to demonstrate the efficacy of the proposed block convolution on both the algorithm side and the hardware side. Specifically, we evaluate block convolution on 1) VGG-16, ResNet-18, ResNet-50, and MobileNet-V1 for ImageNet classification task; 2) SSD, FPN for COCO object detection task, and 3) VDSR for Set5 single image super-resolution task. Experimental results demonstrate that comparable or higher accuracy can be achieved with block convolution. We also showcase two CNN accelerators via algorithm/hardware co-design based on block convolution on memory-limited FPGAs, and evaluation shows that both accelerators substantially outperform the baseline without off-chip transfer of intermediate feature maps." @default.
- W3163130003 created "2021-05-24" @default.
- W3163130003 creator A5001982715 @default.
- W3163130003 creator A5045396120 @default.
- W3163130003 creator A5065157677 @default.
- W3163130003 creator A5074942308 @default.
- W3163130003 date "2021-05-19" @default.
- W3163130003 modified "2023-09-23" @default.
- W3163130003 title "Block Convolution: Towards Memory-Efficient Inference of Large-Scale CNNs on FPGA" @default.
- W3163130003 cites W2097117768 @default.
- W3163130003 cites W2194775991 @default.
- W3163130003 cites W2242218935 @default.
- W3163130003 cites W2276486856 @default.
- W3163130003 cites W2294282016 @default.
- W3163130003 cites W2522548197 @default.
- W3163130003 cites W2525098248 @default.
- W3163130003 cites W2565639579 @default.
- W3163130003 cites W2584311934 @default.
- W3163130003 cites W2584616277 @default.
- W3163130003 cites W2612445135 @default.
- W3163130003 cites W2896983500 @default.
- W3163130003 cites W2963674932 @default.
- W3163130003 cites W2964299589 @default.
- W3163130003 cites W2979455536 @default.
- W3163130003 cites W2980104813 @default.
- W3163130003 cites W2981972129 @default.
- W3163130003 doi "https://doi.org/10.48550/arxiv.2105.08937" @default.
- W3163130003 hasPublicationYear "2021" @default.
- W3163130003 type Work @default.
- W3163130003 sameAs 3163130003 @default.
- W3163130003 citedByCount "0" @default.
- W3163130003 crossrefType "posted-content" @default.
- W3163130003 hasAuthorship W3163130003A5001982715 @default.
- W3163130003 hasAuthorship W3163130003A5045396120 @default.
- W3163130003 hasAuthorship W3163130003A5065157677 @default.
- W3163130003 hasAuthorship W3163130003A5074942308 @default.
- W3163130003 hasBestOaLocation W31631300031 @default.
- W3163130003 hasConcept C11413529 @default.
- W3163130003 hasConcept C114614502 @default.
- W3163130003 hasConcept C138885662 @default.
- W3163130003 hasConcept C153180895 @default.
- W3163130003 hasConcept C154945302 @default.
- W3163130003 hasConcept C173608175 @default.
- W3163130003 hasConcept C2524010 @default.
- W3163130003 hasConcept C2776401178 @default.
- W3163130003 hasConcept C2777210771 @default.
- W3163130003 hasConcept C33923547 @default.
- W3163130003 hasConcept C41008148 @default.
- W3163130003 hasConcept C41895202 @default.
- W3163130003 hasConcept C42935608 @default.
- W3163130003 hasConcept C45347329 @default.
- W3163130003 hasConcept C459310 @default.
- W3163130003 hasConcept C50644808 @default.
- W3163130003 hasConcept C7366592 @default.
- W3163130003 hasConcept C74193536 @default.
- W3163130003 hasConcept C81363708 @default.
- W3163130003 hasConcept C9390403 @default.
- W3163130003 hasConceptScore W3163130003C11413529 @default.
- W3163130003 hasConceptScore W3163130003C114614502 @default.
- W3163130003 hasConceptScore W3163130003C138885662 @default.
- W3163130003 hasConceptScore W3163130003C153180895 @default.
- W3163130003 hasConceptScore W3163130003C154945302 @default.
- W3163130003 hasConceptScore W3163130003C173608175 @default.
- W3163130003 hasConceptScore W3163130003C2524010 @default.
- W3163130003 hasConceptScore W3163130003C2776401178 @default.
- W3163130003 hasConceptScore W3163130003C2777210771 @default.
- W3163130003 hasConceptScore W3163130003C33923547 @default.
- W3163130003 hasConceptScore W3163130003C41008148 @default.
- W3163130003 hasConceptScore W3163130003C41895202 @default.
- W3163130003 hasConceptScore W3163130003C42935608 @default.
- W3163130003 hasConceptScore W3163130003C45347329 @default.
- W3163130003 hasConceptScore W3163130003C459310 @default.
- W3163130003 hasConceptScore W3163130003C50644808 @default.
- W3163130003 hasConceptScore W3163130003C7366592 @default.
- W3163130003 hasConceptScore W3163130003C74193536 @default.
- W3163130003 hasConceptScore W3163130003C81363708 @default.
- W3163130003 hasConceptScore W3163130003C9390403 @default.
- W3163130003 hasLocation W31631300031 @default.
- W3163130003 hasOpenAccess W3163130003 @default.
- W3163130003 hasPrimaryLocation W31631300031 @default.
- W3163130003 hasRelatedWork W2295021132 @default.
- W3163130003 hasRelatedWork W2778739249 @default.
- W3163130003 hasRelatedWork W2804490572 @default.
- W3163130003 hasRelatedWork W2949649078 @default.
- W3163130003 hasRelatedWork W2953116695 @default.
- W3163130003 hasRelatedWork W2971742398 @default.
- W3163130003 hasRelatedWork W3106036237 @default.
- W3163130003 hasRelatedWork W3159557112 @default.
- W3163130003 hasRelatedWork W3176169512 @default.
- W3163130003 hasRelatedWork W3101881194 @default.
- W3163130003 isParatext "false" @default.
- W3163130003 isRetracted "false" @default.
- W3163130003 magId "3163130003" @default.
- W3163130003 workType "article" @default.