Matches in SemOpenAlex for { <https://semopenalex.org/work/W4317526202> ?p ?o ?g. }
Showing items 1 to 79 of
79
with 100 items per page.
- W4317526202 endingPage "151" @default.
- W4317526202 startingPage "137" @default.
- W4317526202 abstract "To improve the inference accuracy of neural networks, their size and complexity are growing rapidly, making the deployment of complex task models on mobile devices with efficient inference a major challenge for industry today. Low-precision quantization is one of the key methods to achieve efficient inference on complex networks, but previous works often quantize partial layers because severe accuracy degradation occurs when quantizing is applied to the entire network. In order to improve the stability and accuracy of low-precision quantization-fine-tuning, we propose a hardware-friendly low-precision full quantization method, called DRGS, which dynamically selects rounding mode for weights according to the direction of weight updates during the training forward and scales the corresponding gradient, finally completing the quantization of all layers of the complex network to achieve floating-free-inference. To validate the effectiveness of DRGS, we apply it to RetinaNet with full 4-bit quantization, and the result of the MS-COCO dataset shows that DRGS has a 2.1% improvement in mAP or at least 2X less quantization loss compared to the state of art implementation. This improvement is also significant even on the YOLO, an object detection model family known for run-time low latency and efficiency. In the latest version of YOLO-v5s, the 4-bit fully quantized network reaches mAP 33.4 which to our knowledge is the best mAP achieved at this category." @default.
- W4317526202 created "2023-01-20" @default.
- W4317526202 creator A5018452576 @default.
- W4317526202 creator A5033886915 @default.
- W4317526202 creator A5050649009 @default.
- W4317526202 creator A5064736086 @default.
- W4317526202 date "2022-01-01" @default.
- W4317526202 modified "2023-10-18" @default.
- W4317526202 title "DRGS: Low-Precision Full Quantization of Deep Neural Network with Dynamic Rounding and Gradient Scaling for Object Detection" @default.
- W4317526202 cites W1861492603 @default.
- W4317526202 cites W2799404913 @default.
- W4317526202 cites W2884150179 @default.
- W4317526202 cites W2962298324 @default.
- W4317526202 cites W2963037989 @default.
- W4317526202 cites W2963122961 @default.
- W4317526202 cites W2963232127 @default.
- W4317526202 cites W2963351448 @default.
- W4317526202 cites W2982644126 @default.
- W4317526202 cites W3004061291 @default.
- W4317526202 cites W3035183452 @default.
- W4317526202 cites W3166947214 @default.
- W4317526202 cites W3205142155 @default.
- W4317526202 cites W3216926558 @default.
- W4317526202 doi "https://doi.org/10.1007/978-981-19-9297-1_11" @default.
- W4317526202 hasPublicationYear "2022" @default.
- W4317526202 type Work @default.
- W4317526202 citedByCount "1" @default.
- W4317526202 countsByYear W43175262022023 @default.
- W4317526202 crossrefType "book-chapter" @default.
- W4317526202 hasAuthorship W4317526202A5018452576 @default.
- W4317526202 hasAuthorship W4317526202A5033886915 @default.
- W4317526202 hasAuthorship W4317526202A5050649009 @default.
- W4317526202 hasAuthorship W4317526202A5064736086 @default.
- W4317526202 hasConcept C111919701 @default.
- W4317526202 hasConcept C113775141 @default.
- W4317526202 hasConcept C11413529 @default.
- W4317526202 hasConcept C136625980 @default.
- W4317526202 hasConcept C154945302 @default.
- W4317526202 hasConcept C2524010 @default.
- W4317526202 hasConcept C2776214188 @default.
- W4317526202 hasConcept C28855332 @default.
- W4317526202 hasConcept C33923547 @default.
- W4317526202 hasConcept C41008148 @default.
- W4317526202 hasConcept C50644808 @default.
- W4317526202 hasConcept C76155785 @default.
- W4317526202 hasConcept C82876162 @default.
- W4317526202 hasConcept C99844830 @default.
- W4317526202 hasConceptScore W4317526202C111919701 @default.
- W4317526202 hasConceptScore W4317526202C113775141 @default.
- W4317526202 hasConceptScore W4317526202C11413529 @default.
- W4317526202 hasConceptScore W4317526202C136625980 @default.
- W4317526202 hasConceptScore W4317526202C154945302 @default.
- W4317526202 hasConceptScore W4317526202C2524010 @default.
- W4317526202 hasConceptScore W4317526202C2776214188 @default.
- W4317526202 hasConceptScore W4317526202C28855332 @default.
- W4317526202 hasConceptScore W4317526202C33923547 @default.
- W4317526202 hasConceptScore W4317526202C41008148 @default.
- W4317526202 hasConceptScore W4317526202C50644808 @default.
- W4317526202 hasConceptScore W4317526202C76155785 @default.
- W4317526202 hasConceptScore W4317526202C82876162 @default.
- W4317526202 hasConceptScore W4317526202C99844830 @default.
- W4317526202 hasLocation W43175262021 @default.
- W4317526202 hasOpenAccess W4317526202 @default.
- W4317526202 hasPrimaryLocation W43175262021 @default.
- W4317526202 hasRelatedWork W2141427150 @default.
- W4317526202 hasRelatedWork W2909808664 @default.
- W4317526202 hasRelatedWork W2945739951 @default.
- W4317526202 hasRelatedWork W2979915190 @default.
- W4317526202 hasRelatedWork W3191026992 @default.
- W4317526202 hasRelatedWork W4213041209 @default.
- W4317526202 hasRelatedWork W4283751495 @default.
- W4317526202 hasRelatedWork W4294982680 @default.
- W4317526202 hasRelatedWork W4317526202 @default.
- W4317526202 hasRelatedWork W4365794625 @default.
- W4317526202 isParatext "false" @default.
- W4317526202 isRetracted "false" @default.
- W4317526202 workType "book-chapter" @default.