Matches in SemOpenAlex for { <https://semopenalex.org/work/W3190259097> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W3190259097 abstract "Skeleton-based Graph Convolutional Networks (GCNs) models for action recognition have achieved excellent prediction accuracy in the field. However, limited by large model and computation complexity, GCNs for action recognition like 2s-AGCN have insufficient power-efficiency and throughput on GPU. Thus, the demand of model reduction and hardware acceleration for low-power GCNs action recognition application becomes continuously higher. To address challenges above, this paper proposes a runtime sparse feature compress accelerator with hybrid pruning method: RFC-HyPGCN. First, this method skips both graph and spatial convolution workloads by reorganizing the multiplication order. Following spatial convolution workloads channel-pruning dataflow, a coarse-grained pruning method on temporal filters is designed, together with sampling-like fine-grained pruning on time dimension. Later, we come up with an architecture where all convolutional layers are mapped on chip to pursue high throughput. To further reduce storage resource utilization, online sparse feature compress format is put forward. Features are divided and encoded into several banks according to presented format, then bank storage is split into depth-variable mini-banks. Furthermore, this work applies quantization, input-skipping and intra-PE dynamic data scheduling to accelerate the model. In experiments, proposed pruning method is conducted on 2s-AGCN, acquiring 3.0x-8.4x model compression ratio and 73.20% graph-skipping efficiency with balancing weight pruning. Implemented on Xilinx XCKU-115 FPGA, the proposed architecture has the peak performance of 1142 GOP/s and achieves up to 9.19x and 3.91x speedup over high-end GPU NVIDIA 2080Ti and NVIDIA V100, respectively. Compared with latest accelerator for action recognition GCNs models, our design reaches 22.9x speedup and 28.93% improvement on DSP efficiency." @default.
- W3190259097 created "2021-08-16" @default.
- W3190259097 creator A5000729469 @default.
- W3190259097 creator A5006683448 @default.
- W3190259097 creator A5039679140 @default.
- W3190259097 creator A5051680867 @default.
- W3190259097 creator A5069310600 @default.
- W3190259097 creator A5075445800 @default.
- W3190259097 creator A5088703563 @default.
- W3190259097 date "2021-08-02" @default.
- W3190259097 modified "2023-10-12" @default.
- W3190259097 title "RFC-HyPGCN: A Runtime Sparse Feature Compress Accelerator for Skeleton-Based GCNs Action Recognition Model with Hybrid Pruning" @default.
- W3190259097 cites W2342311830 @default.
- W3190259097 cites W2619947201 @default.
- W3190259097 cites W2904378456 @default.
- W3190259097 cites W2948058585 @default.
- W3190259097 cites W2981898874 @default.
- W3190259097 cites W2998342322 @default.
- W3190259097 cites W3004884952 @default.
- W3190259097 cites W3035050855 @default.
- W3190259097 cites W3035442541 @default.
- W3190259097 cites W3046757167 @default.
- W3190259097 cites W3105753905 @default.
- W3190259097 doi "https://doi.org/10.48550/arxiv.2108.01020" @default.
- W3190259097 hasPublicationYear "2021" @default.
- W3190259097 type Work @default.
- W3190259097 sameAs 3190259097 @default.
- W3190259097 citedByCount "0" @default.
- W3190259097 crossrefType "posted-content" @default.
- W3190259097 hasAuthorship W3190259097A5000729469 @default.
- W3190259097 hasAuthorship W3190259097A5006683448 @default.
- W3190259097 hasAuthorship W3190259097A5039679140 @default.
- W3190259097 hasAuthorship W3190259097A5051680867 @default.
- W3190259097 hasAuthorship W3190259097A5069310600 @default.
- W3190259097 hasAuthorship W3190259097A5075445800 @default.
- W3190259097 hasAuthorship W3190259097A5088703563 @default.
- W3190259097 hasBestOaLocation W31902590971 @default.
- W3190259097 hasConcept C108010975 @default.
- W3190259097 hasConcept C111335779 @default.
- W3190259097 hasConcept C113775141 @default.
- W3190259097 hasConcept C11413529 @default.
- W3190259097 hasConcept C132525143 @default.
- W3190259097 hasConcept C153180895 @default.
- W3190259097 hasConcept C154945302 @default.
- W3190259097 hasConcept C173608175 @default.
- W3190259097 hasConcept C2524010 @default.
- W3190259097 hasConcept C2778119891 @default.
- W3190259097 hasConcept C28855332 @default.
- W3190259097 hasConcept C33923547 @default.
- W3190259097 hasConcept C41008148 @default.
- W3190259097 hasConcept C42935608 @default.
- W3190259097 hasConcept C45374587 @default.
- W3190259097 hasConcept C6557445 @default.
- W3190259097 hasConcept C68339613 @default.
- W3190259097 hasConcept C80444323 @default.
- W3190259097 hasConcept C86803240 @default.
- W3190259097 hasConcept C9390403 @default.
- W3190259097 hasConceptScore W3190259097C108010975 @default.
- W3190259097 hasConceptScore W3190259097C111335779 @default.
- W3190259097 hasConceptScore W3190259097C113775141 @default.
- W3190259097 hasConceptScore W3190259097C11413529 @default.
- W3190259097 hasConceptScore W3190259097C132525143 @default.
- W3190259097 hasConceptScore W3190259097C153180895 @default.
- W3190259097 hasConceptScore W3190259097C154945302 @default.
- W3190259097 hasConceptScore W3190259097C173608175 @default.
- W3190259097 hasConceptScore W3190259097C2524010 @default.
- W3190259097 hasConceptScore W3190259097C2778119891 @default.
- W3190259097 hasConceptScore W3190259097C28855332 @default.
- W3190259097 hasConceptScore W3190259097C33923547 @default.
- W3190259097 hasConceptScore W3190259097C41008148 @default.
- W3190259097 hasConceptScore W3190259097C42935608 @default.
- W3190259097 hasConceptScore W3190259097C45374587 @default.
- W3190259097 hasConceptScore W3190259097C6557445 @default.
- W3190259097 hasConceptScore W3190259097C68339613 @default.
- W3190259097 hasConceptScore W3190259097C80444323 @default.
- W3190259097 hasConceptScore W3190259097C86803240 @default.
- W3190259097 hasConceptScore W3190259097C9390403 @default.
- W3190259097 hasLocation W31902590971 @default.
- W3190259097 hasOpenAccess W3190259097 @default.
- W3190259097 hasPrimaryLocation W31902590971 @default.
- W3190259097 hasRelatedWork W108745714 @default.
- W3190259097 hasRelatedWork W156843270 @default.
- W3190259097 hasRelatedWork W2007449167 @default.
- W3190259097 hasRelatedWork W2045808965 @default.
- W3190259097 hasRelatedWork W2095778784 @default.
- W3190259097 hasRelatedWork W2143593481 @default.
- W3190259097 hasRelatedWork W2397003076 @default.
- W3190259097 hasRelatedWork W2419277311 @default.
- W3190259097 hasRelatedWork W2582864838 @default.
- W3190259097 hasRelatedWork W2599858257 @default.
- W3190259097 isParatext "false" @default.
- W3190259097 isRetracted "false" @default.
- W3190259097 magId "3190259097" @default.
- W3190259097 workType "article" @default.